This view needs its scripts to run and they have not. Download the file and open the downloaded copy in Chrome or Edge; a Drive, email or Teams preview shows the landing page only.
SIR Management Operating Cockpitv1.0.0
data through 2026-08 · trustee package Nov 10, 2026 Data and run details
Governed sources: SIR Main Workbook · Marketing Calendar ×3 · Industry Benchmark · Trustee Package · data checks 39 pass · 1 warn · 0 fail · 3 skip · run run_34b8a7003aac42c5 · generated 2026-11-16 11:43 EST · engine sir-live-api @ 5081fb6
Populations: 57 restaurants trading in 2026-08 on the operating basis (all brands, pool or not) · 52 in the royalty pool in 2026-08 (pooled basis, roll-ins dated) · 57 in the restaurant-level forecast (every restaurant the forward line projects) · 58 flagged pool-eligible on the roster · 62 geocoded locations of 63 on the roster, closed ones included (weather and site reads) · 53 with meal-period data (day-part reads) · named restaurant by restaurant under Notes and Requirements → Restaurant populations
So what: take out what the calendar did before reading a change as the business moving (an extra week in a 53-week year, the shift that follows it, a month's mix of weekdays and length); what is left is the real move, split into price, guests, openings, closures and a remainder. Pick one window, or a From–To range that sums them; click a row of the table to select it.
Notes
Fiscal period. The labelled comparison (this period against the same-numbered period a year earlier) differs from the like-for-like one the guest basis makes (each traded day against its matched prior day; the same weekday 52 weeks back after a 53-week year) by exactly the calendar: the days traded this period with no prior match and the shift between the labelled prior window and the aligned one. That difference is the calendar effect, in two parts: days without a prior match (the 53-week year's fifth week, and any day with no prior-year twin, as in the periods whose aligned prior falls in the cyber outage, where the two parts largely offset) and the prior window shift (the aligned prior window against the labelled one). The aligned like-for-like change is the underlying move, split into the guest basis's price and traffic effects and a like-for-like remainder. Calendar month. The same restaurants, months and classifications the Revenue bridge sums (comparable, new, closed, other; the Royalty Pool scope by the pooled basis's membership), so the reported change, openings, closures and remainder equal the bridge's and the calendar split sits inside its existing-restaurant change: the prior month recomposed to this month's weekday counts at the prior month's own revenue per weekday is the calendar-equivalent base; recomposed minus prior is the calendar effect (day count plus weekday mix), current minus recomposed the underlying like-for-like move. The fiscal grain follows the guest basis's pool flag instead, so its Royalty Pool restaurant set can differ from the month grain's. Against the window before. The prior year's own move between the same two windows, on the current windows' day counts, is the seasonal norm; what is left is the change in like-for-like growth. Openings, closures and a store-comparability remainder close every bridge exactly (data invariant calendar_baseline_identity). Holiday placement is not part of the calendar effect in this version; a holiday that moves between windows stays in the underlying move. The Operating / Royalty Pool switch in the header sets the scope.
Reported change—vs the same window a year earlier
Calendar effect—what the calendar alone moved
Underlying change—calendar removed
Openings & closures—from openings and closures
Remainder to allocate—store-comparability remainder · detail in The gap
Read—
Reported change vs underlying change, by window
Where the two lines part, the calendar made the difference.
Notes
Each point is one window's change against the same window a year earlier, as a share of the prior base: reported, and with the calendar effect removed.
ReportedUnderlying (calendar removed)
Bridge — calendar first, then the rest
Notes
The reported change for the selected window, calendar first: its two parts, then the like-for-like move split into price and traffic (fiscal grain) or shown whole (calendar month), then openings, closures and the remainder. Every segment reconciles to the closing total.
Across the range
Notes
Running totals across the selected range: the reported change, the calendar effect inside it and the underlying change left once the calendar is removed. Every window's bridge closes exactly, so the range sums do too.
The known causes that overlap the selected window, with the days they cover in this window, in the same window a year earlier and in the window before: recorded public and business events (the ledger; no figure of their own, the size is read from the window's own price and traffic effects), holiday placement (each holiday's revenue against its ordinary expectation, this window against a year earlier, brand level), local events near the restaurants (the supported effect estimates times the matched restaurant-days), closures from the renovation register (closed days times baseline daily revenue), revenue shifts the monitor found, and the weather lens at the period grain. The measured causes sum to the explained change; what is left of the like-for-like change is shown as unexplained, never forced to zero.
Cause
This window
A year earlier
Window before
Effect
Basis
Note
Against the window before
Notes
Like for like on the restaurants comparable in both windows: the change against the window before, the seasonal norm (the prior year's own move between the same two windows, on the current windows' day counts) and the underlying change left once the norm is removed.
Change vs the window beforeSeasonal normUnderlying
Every window
Show as table
Window
Reported
Calendar
Underlying
Like-for-like
Price
Traffic
Openings
Closures
Remainder
Read
vs window before
Seasonal norm
Underlying
Read
Remainder to allocate — by brand
So what: the remainder sits with restaurants that were not open in both windows; it is not a data gap.
Notes
Each brand's remainder over the selected window or range, on the same grain and scope, as a share of its prior base; the largest window names where it concentrated. The finer, evidence-tested breakdown is in The gap (Notes & Requirements).
Brand
Remainder (CAD)
Share of prior base
Largest window
Explaining the change
All-restaurant
All periods = cumulative gap; a single period = single-period view — the same choice as the Revenue bridge's range picker.
So what: how much of the reported revenue change independent evidence can account for — guest traffic, spend per guest, the new-restaurant ramp, a prior-year correction — and where the part it can't yet explain sits. The Bottom line below is the read; the sections under it are the working.
Notes
The revenue change that comparable-restaurant performance and openings/closings don't explain, tested against real evidence — guest traffic, spend per guest, weather, calendar — before it's trusted. What passes joins the explained baseline; what doesn't stays open, and shouldn’t drive distribution planning yet. Guest traffic and spend per guest are shown as signed net figures across restaurants — restaurants above and below expectation partly offset, so they aren’t meant to sum to the supported total, and no movement is counted twice.
What "expectation" means, and how to read these flags
Throughout this tab a restaurant is compared with its expectation — a fair-value estimate of what it should have earned this period, built from its own recent history and the period's calendar, with a calibrated range around it. It is not a budget or a target. A restaurant reads above or below expectation only when the actual lands outside that range — a genuine surprise, not normal period-to-period wobble — and gap size is how far outside, measured against that restaurant's own normal swing. Read these as a triage lens, not a governed number: a gap that repeats across periods is structural and worth investigating; a one-off is usually noise. Nothing here adjusts distributions or proves a cause — it points you to the restaurant, and the drivers (guest traffic, spend per guest, mix) explain the why. Full derivation on the How this is calculated page.
How this compares to the Reported revenue change tabs
A wider lens, not the same number. On the Reported revenue change tabs, "remainder to allocate" is the small leftover of the year-over-year change — what's left after same-restaurant growth and openings/closings. Here, every restaurant-period is measured against a fair expectation of what it should have earned: a different, usually larger baseline, so the two use different yardsticks and aren't meant to match. They also label time differently — the Reported revenue change tabs by calendar month (e.g. 2026-04), this tab by SIR fiscal period (e.g. FY2026 P9).
Starting revenue gap—Unexplained revenue, before the second review
Revenue now supported—Explained by evidence · no double counting
Share of gap supported—How much the revenue gap fell
Open revenue gap—Still unexplained · next investigation priority
How the gap narrows under evidence
So what: one lever does almost all the work: guest traffic and spend per guest. The internal-pattern and external tests add next to nothing on top. The final bar is what no evidence yet explains.
Notes
Read top to bottom. Each dark bar is the gap that still had no explanation after that test; the green to its right is how much that test just explained. A short final bar means most of the starting gap is now accounted for — a long one means real questions remain.
Gap still unexplainedExplained by this testStill open
Where the open gap concentrates
So what: the unexplained money is not spread evenly across periods; the restaurant-periods at the top of the list are where an explanation changes the read most.
Largest remaining gaps
So what: the biggest still-open store-periods are the ones to look into: how large each gap is, how far it runs from expectation, and which way. Persistent gaps below expectation in one restaurant are the priority; windfalls (above) are the upside surprises.
Notes
These are the first items where additional evidence can still create value. Select one above to see its story in the chart.
Restaurant
Period
Remaining gap (CAD)
% vs expected
Direction
Patterns over time
So what: what tends to come before a gap, and how each kind of gap usually resolves the period after.
Notes
Ordered co-occurrences across consecutive periods, and the base rate at which each gap pattern recovers, persists or worsens next period. Ordered patterns, not causes.
Investigation canvas
So what: pick a flagged case and follow it: the observed variance, what was nearby, what moved together, what happened before, and what to do next.
Notes
The investigation workspace: it connects the store-period residual cases so an analyst can move from a flagged gap to related cases and possible explanations — none of it changes the forecast. Investigation relevance, not causality.
Royalty Pool basis
So what: management’s budget is accurate at the pool total and uneven restaurant by restaurant, so the restaurants furthest from plan are the review priority and a ⟳ gap (a closure or an opening the plan could not see) is expected, not a forecast error. Management’s budget against reported FS revenue, and which restaurants drive the difference.
Notes
SIR management's own budget against reported financial-statement (FS) revenue — the figure the fund reports on — per fiscal period. Where actual diverges from plan, and which restaurants drive it.
Evaluation year—Fiscal periods with budget and actual
Pool total vs plan—Actual revenue vs budget · evaluation year
So what: the restaurants furthest from plan are the review priority; a ⟳ dot had a closure or opening the plan could not see.
Notes
Each dot is one royalty-pool restaurant's variance to management's budget over the selected scope. Dots right of the line came in over plan, dots left came in under — the furthest are the review priority. Stores marked ⟳ reno/new had a major closure inside this year or opened mid-year, so a plan anchored to last year can't see them — those gaps are expected. Minor or years-past closures are not flagged: the renovation model shows stores recover fully within a week or two, so those variances are genuine. The unflagged stores are the operational variance. The full ranked list, with click-through to period detail, is below.
Restaurant
Brand
Budget (CAD)
Actual (CAD)
Variance (CAD)
Variance %
Plan vs actual by brand — all seven at once
So what: each brand on its own scale shows where actual and budget part; a brand whose lines part in the same periods as the pool total is where the plan’s assumption was wrong.
Notes
Actual vs budget for each brand, without changing the brand filter. Each card is zoomed to its own revenue range so the gap to plan is visible; the % badge gives the size of the variance, so read that to compare brands rather than one card's height against another's. Teal = reported actual · navy = management budget.
Budget vs actual by period
So what: the gap between actual and budget at the pool total is what reaches the year-end true-up.
Notes
Both series are FS net revenue for the royalty-pool restaurants, summed across the royalty pool by fiscal period, labelled at each line. Where actual sits above budget, the pool total came in above plan. The vertical axis is zoomed to the revenue band — it does not start at zero — so the gap between plan and actual is legible; size the variance from the axis values, not the visual gap alone.
Accurate pool total, uneven store-level plan
So what: the pool total lands close to plan because restaurant variances offset each other; sharing management’s total out by last year’s restaurant shares keeps that total and cuts the restaurant-level variance.
Notes
Errors measured as WAPE (weighted absolute percentage error). The aggregation staircase shows how restaurant-level variances shrink as they roll up to brand and pool total. The reconciled split holds the governed pool total fixed and re-distributes it by the standalone line's store shares (the line is named on the panel), so the pool total the fund reports is unchanged while each restaurant's plan gets sharper.
Price & traffic
Operating basis
So what: for the brand chosen, is revenue growing on more guests or on a higher cheque? Price, cheque, discount and margin are below; the read to watch is a cheque rising while guests fall. The meal-period, channel and department mix is on Sales mix. Restaurant by restaurant, the guest-change against check-change map is on Store Breakdown → Restaurant drivers.
Notes
Comparable-store revenue split into guest traffic and average check for the brand picked here; the one-dot-per-restaurant map is on Restaurant drivers. The pattern to watch: revenue held up by price increases while guests quietly fall away. The average check blends all channels; a shift toward delivery can read as a price move — see Notes & Requirements → Method notes.
Pricing, discount & margin
Operating basis
So what: a cheque move is split into a menu-price step and mix; a uniform step down across departments is discounting.
Notes
The comparable-store price effect — the part of revenue that came from spend-per-guest, not headcount — split into a menu-price step (the whole check moving together) and mix (guests spending differently across bar, dining, patio, take-out and delivery). A uniform step down is the signature of discounting. Pick a brand.
By brand — price vs discount decomposition
Filtered by the brand selected above.
Menu-price step vs discount, by period
Notes
Each point is the common per-guest move that period — the rate the whole check shifted at, robust to any single department swinging. Above the line is a menu-price step up; below is a step down — the discount / comp signature. Mix sits separately in the readout.
Where the check moved — per-guest change by department, full year
Notes
Change in comparable-store spend per guest by department, this year vs last, in dollars (the percent is in brackets). All bars pointing the same way = a broad menu-price step or a broad discount; bars pointing different ways = mix, guests reallocating their wallet. Dollars, not percent, so a big swing on a tiny department can’t dominate. Take-out (SIR-operated) and delivery (third-party apps) are shown separately; the bars sum to the full-year check change per guest.
These reads span the whole portfolio and are not filtered by the brand selector above.
Price effect — inflation pass-through vs real pricing
So what: menu prices against dining-out inflation say how much of the price effect is real pricing rather than pass-through.
Notes
Portfolio-level (CPI) context. Splits the menu-price contribution to comparable revenue (CAD) into the part explained by dining-out inflation — the market rate menus were rising anyway — and genuine pricing action above it. So what: bars above the line are menu prices rising faster than inflation; below, SIR priced under inflation. Re-reads the price effect against CPI without changing any governed figure.
What is driving promotions
So what: whether the promo rate follows inflation or moves on its own, read by how closely the two move together.
Notes
Portfolio-level. The governed promo-discount rate over time, tested against inflation. A rising line is an effective price cut that offsets menu pricing. The takeaway states whether it tracks inflation (a defensive reflex) or trends independently (a commercial choice) — measured by correlation, not assumed.
Brand dossier
Operating basis
So what: one brand's trade between guests and price, end to end: whether lower prices brought guests, or a higher cheque held revenue while guests fell.
Notes
One brand’s traffic–price trade, read end to end: is it giving up price to buy guests, or holding revenue with price while guests fall? The verdict adapts to what the numbers show. Pick a brand.
Spend per guest vs guests — indexed to last year
Notes
Both series on one indexed scale (last year = 100), so the gap between the lines is the story — and a widening gap is a bigger bet. Lines pulling apart with guests on top is a discount buying traffic; guests sliding under 100 while spend-per-guest climbs is a higher cheque making up for fewer guests.
Where the revenue change comes from
So what: the comparable change is a guest effect plus a spend-per-guest effect; their sum is the net.
Notes
The comparable revenue change splits exactly into a guest-traffic effect and a spend-per-guest (price) effect. Their sum is the net.
Whether the trade paid, by period
Notes
Net comparable revenue change each period; the multiplier beneath each bar is that period’s exchange — guests gained per 1% of check given up. Red bars are periods where the trade was net negative, and a multiplier under 1.0 is where volume stopped covering the price given up — watch whether those cluster early (teething) or late (the lever running out).
Where the check moved — spend per guest by department, full year
Notes
Dollars per guest, this year vs last, for the comparable stores. All bars the same way = a broad menu-price step or discount; different ways = mix. Take-out (SIR-operated) and delivery (third-party apps) are shown separately.
Margin lens — revenue vs gross profit
So what: once margin and promo-cost data land, this says whether a discount that grew revenue also grew profit.
Notes
The reads above are revenue. This panel turns them into contribution once category-margin and promotion-cost data are supplied — the one thing that decides whether a discount that grew the top line also grew the bottom line. It activates automatically when the data lands.
Sales mix
Operating basis
So what: where the revenue change comes from for the brand chosen: which meal period, dine-in against off-premise, and which department. The price, cheque, discount and margin reads are on Price & traffic.
Which meal period drives the change
Notes
Comparable restaurants of the selected brand; each fiscal period is compared with the same period days a year earlier and guests sit on the guest basis. Traffic effect = guest change × average spend per guest; spend effect = spend-per-guest change × average guests. Spend per guest is net sales ÷ SIR GuestCount by meal period, not an average cheque.
Fiscal year to date — revenue change by meal period, guests vs spend
Guests by meal period — each period vs last year
Sales mix — where the check is spent
So what: where the cheque is spent, by service channel (dine-in against off-premise) and by POS department, for the brand selected above.
Notes
Real POS department and channel shares for the selected brand, by fiscal period. The dine-in vs off-premise shift shown below is the exact figure the Explain & evidence "Internal pattern test" checks.
Dine-in versus off-premise mix
So what: which way revenue is shifting between dine-in and off-premise; the same figure the internal-pattern test on the “The gap” page checks.
Notes
A shift in revenue mix. Exact attribution needs guests and spend by channel. This is the channel-mix figure tested against the unexplained gap by the internal-pattern test on the “The gap” page.
Show as table
Brand
Period
In-restaurant %
Prior %
Change (pp)
Bar department share—Not alcohol specifically — a POS department
Patio department share—Share of revenue dollars
Take-out share—SIR-operated pickup · fully in SIR's control
Delivery share—Third-party apps · app fees, not SIR-controlled
Periods covered—Real POS department data
Revenue mix by department
So what: the share of the cheque each POS department takes, period by period.
Notes
Real Bar, Dining, Patio, Delivery and Takeout revenue shares for the selected brand, by fiscal period — newly surfaced from raw POS data.
Inflation & market
So what: set against CPI, the question is whether growth is price or guests; the cheque against inflation and guests against last year are below.
Notes
SIR growth and menu pricing set against Statistics Canada CPI (table 18-10-0004) — separating genuine volume growth from price inflation, and showing whether menu pricing tracks dining-out inflation.
What the inflation lens says about the business
So what: how growth is produced, price against volume, and how the CPI benchmark is built. Promotions and inflation pass-through are on Pricing, discount & margin.
Notes
The conclusions, not just the series: whether growth is carried by price or volume, and how the weighted CPI benchmark is constructed. Each is evidenced below.
How revenue is growing — price against volume
Notes
Comparable-revenue movement split into price (spend per guest) and volume (guest traffic) over the available history. A large price bar with a negative volume bar means growth is carried by pricing while customer counts slip — durable only while price can outrun the traffic it costs.
Nominal vs real comparable growth
Notes
SIR comparable-store revenue growth (nominal) deflated by all-items CPI to real growth. The gap between the lines is price inflation; the real line is genuine volume.
Where food-cost inflation is biting — latest month
So what: food categories right of the dining-out line are input costs rising faster than menu prices, which squeezes margin.
Notes
Latest-month CPI by food category, ranked — the grocery costs behind the kitchen. The dashed line is dining-out inflation (what restaurants are charging). So what: categories to the right of the line (amber) are input costs rising faster than menus can — margin pressure; those to the left (teal) still have headroom.
Spend per guest vs guest traffic, by period
So what: where spend per guest rises faster than guests fall, price is carrying the brand; where guests fall faster, it is not. Both year on year on the same period days a year earlier.
Notes
The two forces behind comparable revenue, by fiscal period: spend per guest (average cheque) against guest traffic, both year over year. When the cheque line holds up while traffic bends down, price is holding revenue while volume falls — the tension inflation creates.
Spend per guest vs inflation
So what: the real cheque is the cheque with inflation removed; below zero it rose less than prices did.
Notes
Real average cheque by fiscal period — spend per guest year over year with all-items CPI removed. Bars above the zero line are a genuine per-guest gain; bars below mean the cheque rose but still rose less than inflation. The zero line is inflation.
SIR vs industry
So what: SIR against the market for its provinces; the gap is where SIR sits above or below the market.
Notes
Statistics Canada restaurant-industry benchmark, weighted to the provinces SIR operates in. Only year-on-year movement is shown (absolute industry revenue levels are context, not a governed figure). Additional context for the Revenue bridge and The gap — not one of the two hypotheses tested against the unexplained gap (that's Price & guests and Weather/Calendar). The co-movement panel below correlates SIR against the industry on raw year-on-year over the months clear of the GST/HST tax-holiday base effect — a short overlap. See Notes & Requirements for the basis and the alternatives considered.
Latest industry YoY—Weighted to SIR's operating provinces
Latest SIR YoY—Selected concept
Alpha vs industry—SIR growth minus industry growth
Benchmark coverage—Geographic weight coverage
SIR–industry correlation—2-yr annualized basis
SIR versus industry — year-on-year growth
Notes
Dashed line is the weighted industry benchmark. The amber segment marks the GST/HST tax-holiday months, whose prior-year base is distorted; the teal tax-adjusted line is a modeled estimate of the underlying industry trend across those months — each month compounded with the same month a year earlier and annualized, which divides out the distorted base. Modeled context, not a governed figure. Hover a point for the month's detail.
Relative performance — SIR minus industry (pp)
Notes
Positive bars mean SIR grew faster than the industry that month; negative means SIR grew slower.
Co-movement — does SIR track the market?
Notes
Both axes are raw year-on-year growth. The dashed diagonal is parity (SIR = industry); the solid line is the fitted trend through the clean months, and r is their correlation. The grey dots are the GST/HST tax-holiday months (Dec 2025–Mar 2026): they are plotted for transparency but excluded from the fit, because their prior-year base is distorted. The SIR-to-benchmark overlap is short (SIR's year-on-year starts late-2024), so this is a small sample — directional context, not a significance test. A 2-year annualized basis (to cancel the tax base and keep those months) was tested but collapses to too few points on this overlap; see Notes & Requirements.
New restaurants vs existing
So what: the existing-restaurant line is the rate to plan on; openings add on top of it. Below, the openings against the ramp, the renovations against matched peers, and every closure on record.
Notes
Separates estate change — restaurants opening and closing — from how the restaurants open in both periods performed. Additional context, not one of the two hypotheses tested against the unexplained gap (that's Price & guests and Weather/Calendar) — openings and closures are already accounted for directly on the Revenue bridge.
Restaurants trading—Latest month, operating basis · all brands, in the pool or not
Net change, last 12 months—Net restaurants (openings − closures)
Openings and closures effect, latest month—Share of prior-year revenue
Comparable growth, latest month—Same restaurants, both periods
Operating restaurant count over time
So what: a step down in the count with no closure on record is a data gap, not a closure.
Notes
Restaurants currently reporting sales, by concept. This is operating status, not royalty pool membership — the two are not the same thing.
Openings and closures effect — All brands
So what: openings add revenue the existing restaurants did not earn and closures take it away; the existing-restaurant line is the pool’s momentum, the bar beyond it is built growth.
Notes
Opening revenue minus closure revenue, as a share of prior-year revenue. Separates "grew by opening more restaurants" from "grew at existing restaurants."
New restaurants vs the maturation ramp
So what: how the recent openings are earning in, against what a restaurant of their age usually earns.
Recent openings vs their age-expected ramp
Loading new-restaurant performance…
Renovations against matched peers All-restaurant
So what: what renovations have added so far: every renovated restaurant against matched comparison restaurants over the same months, so the estate’s own trend is taken out — the same kind of pooled read as the openings against the ramp above.
Notes
The portfolio renovation effect estimated against the nearest comparable restaurants of the same brand and size over the same months, so calendar, seasonality and the wider estate trend cancel. Investigation-grade context — it enters the forecast only through the feature tournament's out-of-period gate. One restaurant's own recovery and its reads are on Store Analytics. See How this is calculated.
Every renovated restaurant
So what: every renovated restaurant, how many there are and which are back to their own level.
Notes
Every restaurant with a detected renovation (refresh / major / re-brand). Level vs before = mean reported revenue in the up-to-3 months after reopening against the up-to-3 months before closing; trajectory = first post-reopening month vs latest. Small-sample, single-restaurant reads; the portfolio-level effect with matched comparisons is at the top of this subtab.
Renovated restaurants—With a detected renovation
Median closure length—Across renovated restaurants
At or above prior level—Post-reopening average vs before
Below prior level—Post-reopening average vs before
Post-renovation recovery — all renovated restaurants
So what: each renovated restaurant’s settled level against its own pre-closure base; a restaurant still below its base months after reopening is the one to ask about. Click a row to open it on Store Analytics with its reads.
Notes
Level vs before = mean reported revenue in the up-to-3 months after reopening against the up-to-3 months before closing. Since reopening = first post-reopening month vs latest.
Restaurant
Brand
Class
Closed → reopened
Days
Level vs before
Since reopening
What we detected — closures by length
Notes
SIR has no standalone renovation dataset, so the read is built from the daily operating signal (which already reflects the curated "0 Sales Days" closure log). A run of consecutive closed days is classified by length — a short closure (3–6 days, cause unknown — usually not a renovation), refresh (7–10 days), major renovation (11–20), re-brand (over 20 days with a concept change) and extended major (over 20 days, no concept change); closures under 3 days are excluded. Only the renovation classes feed the recovery read below. Where a curated closure record covers the run, its reason is carried and the row is marked from closure log; otherwise inferred from gap. Descriptive context and changes no governed figure unless separately promoted through the model's gate.
Closures classified by length
Loading closure classification…
Closure event register
So what: every detected closure of three or more days with its class; a closure not on this register is one the plan and the forecast do not know about.
Notes
Each detected closure of 3 or more days (including short cause-unknown closures), with its class, the concept before and after, the source (curated closure log vs inferred from an operating-day gap), and a clean pre-closure daily-revenue baseline. Descriptive register — investigation-only.
Restaurant
Brand
Region
Closed
Reopened
Days
Class
Concept change
Source
Reason
Pre-closure $/day
Each concept on the public record — the pooled read, the dollars since conversion, the drivers, the test and the register — is on New concepts; each site’s reads are on Store Analytics.
New concepts All-restaurant
So what: is each new concept earning more than the restaurant it replaced would have? The answer is read on the sites’ own figures against matched peers, and the verdict strip says what to do with it.
Notes
One method for every concept: each conversion against its matched peers over the same calendar months (the renovations’ matched control, on this concept’s conversions alone), with the restaurant’s own prior year as the placebo and the leave-one-out. Under four conversions the band is the estate’s own spread — every untreated restaurant “reopened” at every month its data allows and read the same way — so two sites with two months are held to the band two sites and two months can give. An opening is read against its brand’s ramp. Every read is on the restaurant’s total revenue (the concept has no POS line of its own); an evidence-supported association, never a gain, and nothing here enters a forecast.
The words on this page, in plain English
So what: every figure on this page is one of a few reads, and each is defined here once; hover a dotted word anywhere on the page for its meaning.
The reads, with their bands
Notes
Each dot is a read against matched peers; a shaded bar is its 90% band (green when the whole band is above zero, red when below, grey when it spans zero). The pooled read sits on the first months after reopening; “without” rows leave one site out in turn; “beyond the sites’ own prior year” takes out each site’s usual seasonal move; an opening is against its brand’s ramp. Hover a row for the figures behind it.
Month by month since the change — each site against its peers’ path
Notes
Each point is the month’s actual revenue against what the site’s matched peers’ path implies from its pre-closure level, as a percentage; month 1 is the first complete month after reopening. A shaded column is a month inside a known window (a tournament, a tax holiday, an outage); the dashed lines mark the six- and twelve-month reads. The first-months read on the chart above sits on the left; the dollars run over every month; the last point is where the site stands now. Hover a point for the dollars.
Every site in one view
So what: a site’s card says where it stands now against its peers; click it for the full reads on Store Analytics, and hover a figure or a chip for its basis.
Notes
A conversion’s read is this site against its matched peers from the months before the closure to the months after reopening; its dollars are the months since reopening against what its peers’ path implies from its pre-closure level; the small line is that path month by month; an opening is read against its brand’s ramp; plan is management’s budget on the FS basis. A chip names a known window inside the first-months read, an SIR opening, closure or reopening nearby during the read, and the closed days inside the plan year.
The same sites as a table
Restaurant
Kind
From
Read
First months vs peers
Months after reopening
Since the change
Guests vs peers
Cheque vs peers
Vs the opening ramp
Vs plan
Now vs peers
Known window in the post window
SIR change nearby
What moved: more guests, or a bigger cheque?
Notes
Guests per day and spend per guest from the last complete fiscal periods before the closure to the first after reopening, the site’s move minus the same move at its matched peers, on the guest basis. A lift carried by guests is a different proposition from one carried by the cheque. Hover a bar for the site’s own move and its peers’.
How far along each read is
Notes
A site is readable from two complete months after reopening (a month the restaurant traded under 15 days does not count); the six- and twelve-month marks are the firm reads, each with the data month that completes it. The smallest lift the evidence can show is the 90% band’s half-width for the sites read and the shortest post window among them; a concept clears when its band keeps its side of zero with any one site left out. Hover a track for the dates.
Design a test — how many sites, for how long
Notes
Each cell is the smallest lift that many sites read over that many months can tell from zero at 90%: the 90% band’s half-width for the months, over the square root of the sites. The spread behind it is the estate’s own — every untreated restaurant “reopened” at every month its data allows and read as a converted one is — so it carries the estate’s seasonality and noise, not an assumption. Sites are taken as independent; sites that share a trade area or a launch read as fewer. The spread is the estate’s, every brand together, not the concept’s brand alone. Hover a cell for the draws behind it.
The public record behind the sites
Notes
The register is the public record (the Fund’s releases and SIR’s filings): one row per restaurant and concept with the sentence that states it; a row whose sentence is still to be read in the cached release is marked pending. Every read is on the restaurant’s total revenue — the concept has no POS line of its own — so a conversion is read as a closure is (this site against its matched peers over the same months, a single-site read with no band of its own; the portfolio band is above) and an opening against the ramp. Each site’s reads are on Store Analytics under its monthly line; the concept’s review is above. A prototype inside a restaurant needs a POS line of its own before it can be read.
Restaurant
Concept
Kind
City
From
Closure
Status
Source
One restaurant, in depth All-restaurant
So what: one restaurant end to end: its revenue path, what happened to it and how each event read, its meal periods against the brand, what holidays, nearby events and weather do to it, and the shape of its week. Pick the restaurant above. Restaurants side by side are on Restaurant drivers.
Notes
Reported monthly revenue per restaurant (from the store × month reconciliation used across the cockpit), trailing 12 months where available; a partial current month is trimmed so the trend read isn't distorted. A restaurant with a closure, a conversion or an opening on record shows every month and its event reads under the chart (against matched peers, its own prior year, the opening ramp, its plan).
Latest month revenue—Most recent reported month
Trailing-12-month trend—First vs last month in the window
Months of data—Reported months available
Monthly revenue — trailing 12 months
Notes
Reported monthly revenue for the selected restaurant over the trailing year (or all available months if fewer); a restaurant with a closure, a conversion or an opening on record shows every month, the event marked. A partial month is trimmed, and the trend reads complete months only.
What happened at this restaurant
Notes
Each read names its basis. A closure is read against matched peers over the same calendar months (the Openings & closures method on this site alone, no band of its own) and against the restaurant’s own move over the same months a year earlier; an opening against the brand’s ramp; the plan row is management’s budget on the FS basis. The recovery read below compares mean reported revenue in the months just before closing with the months after reopening, and reads the change since reopening (up, flat or down from the first post-reopening month to the latest — a comparison across the whole post-reopening record, not a statement that a restaurant is still ramping). Small-sample, single-restaurant context — the portfolio-level renovation effect with matched comparisons is on Openings & closures.
Read
Figure
Basis
Meal periods — spend per guest and traffic against the brand
Notes
Trailing 52 weeks of POS day-part data for this restaurant. Spend per guest is net sales ÷ SIR GuestCount by meal period (not an average cheque); the index is against the median of the brand’s Royalty Pool restaurants; year-on-year figures compare with the prior 52 weeks at the same restaurant on the guest basis.
Meal period
Spend per guest
Brand median
Index (brand = 100)
Guests per day
Guests YoY
Spend/guest YoY
Revenue YoY
Demand drivers at this restaurant — holidays, local events, weather
Notes
The store-level reads behind the by-restaurant panels on Restaurant drivers, for the restaurant selected above: its holiday uplift over its own normal same-weekday revenue (shrunk toward its catchment type), its uplift on nearby-event days over its matched non-event days, and the weather lens across its fiscal periods (the typical revenue variance on unusual-weather days and the driver named most often). Descriptive context; none of it enters a forecast.
One event day at this restaurant
So what: what one event day did for this restaurant, against similar days without an event.
Notes
Compare the selected event day with similar non-event days.
Revenue gap and weather intensity
So what: whether the days this restaurant sat below its expected level were the unusual-weather days.
Notes
Bars show the daily revenue gap. The line shows weather intensity on the same dates.
Each restaurant’s share of revenue by weekday and meal period over one fiscal year, grouped into a few shapes. The shape is descriptive; its one governed use is as the peer pool for the renovation matched control, where grouping by archetype is tested against grouping by brand (Openings & closures → Renovations against matched peers).
Archetype
Restaurants
Lunch
Dinner
Late
Fri–Sun
Restaurant drivers
Operating basis
So what: every restaurant side by side, so the ones to act on stand out: guest change against cheque change and the net comparable change by restaurant (for the brand selected above), holiday and nearby-event responsiveness store by store, the weather lens on any restaurant-period, and the weather investigation queue. The pool- and brand-level reads stay on Price & traffic and Demand & calendar.
Store map — guest change vs average-check change
So what: the restaurants with fewer guests and a higher cheque (the shaded upper-left corner) are the ones to look at first.
Notes
Each dot is a restaurant: horizontal is its guest change, vertical its average-check change. The shaded upper-left flags restaurants with fewer guests while raising price. Dot size scales with the traffic swing; a ring marks nearby sports-event exposure.
Net comparable revenue change by restaurant
So what: a bar left of the line is a restaurant where the higher cheque did not make up for fewer guests.
Notes
Traffic and price combined. Bars left of the line are restaurants where the price increase did not cover the guest loss — the loss is real, not just masked.
Does a strong rating convert to performance?
So what: a strong rating does not by itself convert to revenue; the restaurants off the diagonal are where rating and performance disagree, and those are the ones to look at. Each restaurant's guest rating against how its revenue compares to its own model expectation — up and right is rated well and above its expected level; the watch list sits bottom left. The vs-expectation axis is where the rating signal shows (raw growth doesn't).
Notes
Each restaurant's Google rating (horizontal) against mean revenue variance versus its own model expectation (vertical), coloured by brand. The horizontal line is meeting expectation; the vertical line is the median rating, splitting the field into four quadrants. A single snapshot — association, not trend or cause; the rating spread is narrow (4.2–4.9). Raw comparable growth shows no link (correlation +0.09) because ratings are bunched and growth is dominated by size and brand; against each restaurant’s own expectation the link is modestly positive (+0.29).
Holiday effect, store by store
So what: a holiday’s effect is local, so each restaurant gets its own read: its responsiveness is pooled across holidays, shrunk toward its catchment type, and checked on a past year kept for testing where the history allows.
Notes
A statutory holiday is local — an office-core restaurant empties while a destination fills. Each store's holiday responsiveness is pooled across every holiday it has seen (not one named holiday, which no single store sees often enough to test), shrunk toward its catchment type, and — where it has the history — checked on a past year kept for testing. Catchment types are relative rankings within the portfolio — thirds of the weekend-vs-weekday revenue mix — since nearly every restaurant does more on weekends.
Holiday responsiveness by store
Notes
Each dot is a store's shrunk holiday-day uplift over its own normal same-weekday revenue; the whisker is its 90% range. Green sits above a normal day, red below; the label shows each store's catchment type. The spread across dots is what one brand-level number hides.
Event effect, store by store
So what: whether a restaurant gains from a nearby event depends on where it sits, not on its brand; pooled across nearby-event days and checked on a past year kept for testing where possible.
Notes
Whether a restaurant captures spillover from a nearby venue is a property of its catchment, not its brand. The table above tests one event type at a time (brand-level, because one restaurant rarely has enough of a single type); here each store's responsiveness is pooled across every nearby-event day it has seen (all event types, ≤5km), shrunk toward its catchment type, and — where it has the history — checked on a past year kept for testing. Same catchment types as Calendar. Pick a restaurant in the filter above to see its own read.
Event responsiveness by store
Notes
Each dot is a store's shrunk uplift on nearby-event days over its own matched non-event days; the whisker is its 90% range. Green sits above a normal day, red below; the label shows each store's catchment type.
Upcoming events near your restaurants
So what: the events ahead, each with what history says it adds, tested or directional.
Notes
Start from normal revenue. Add only real historical event evidence — validated (green) or directional-only (amber), never blank when evidence exists. End with a measurable local plan.
Source coverage. Public event feeds drive the analysis. Optional ticketing feeds are used only when access and client-use approval are configured. A Modeled estimate badge means the uplift is inferred from distance to the venue and its capacity rather than measured on this event type — a directional planning read, not a forecast input.
Weather, restaurant by restaurant
So what: the restaurant-periods whose revenue gap unusual weather can explain, largest first. One restaurant's daily timeline and weather fingerprint are on Store Analytics.
Notes
Compare unusual weather with similar days. Keep it only when the pattern repeats. One of the two hypotheses tested against the open gap on The gap — Calendar (holidays) is the other; see How this is calculated for the full method.
Weather — the past-year test is described in How this is calculated.
Days with weather—Coverage of sales days
Comparable days0Same restaurant and weekday
Patterns that repeated0Worked on the test year
Gaps worth investigating0Weather and gap point the same way
Weather investigation queue
So what: the queue is ranked by the size of the gap and how strongly the weather lines up with it.
Notes
So what: the queue is ranked by the size of the gap and how strongly the weather lines up with it.
Restaurant
Period
Revenue gap
Weather signature
Matched days
Historical range
Demand & calendar
All-restaurant context
So what: what moves demand beyond price and guests, at pool and brand level: holidays against the ordinary same weekday, the shape of the week and the season, local events with their past-year test, and the weather patterns that repeat. Restaurant by restaurant, the same reads are on Store Breakdown → Restaurant drivers.
Notes
Compare like with like. Adjust for weekdays and supported holidays before looking for other explanations. The other of the two hypotheses tested against the open gap on The gap — Weather is the other. Upcoming holidays use each holiday's calendar-adjusted baseline — descriptive context already in the day-of-week reference, not a separate dollar input.
Holiday observations0Restaurant-days evaluated
Supported effects0Repeated history
New discoveries0Passed the past-year test
Upcoming opportunities0Next calendar horizon
Ordinary day versus holiday-adjusted expectation
So what: how much a holiday adds to, or takes from, an ordinary same weekday.
Notes
Observed revenue is compared with an ordinary same-weekday reference and a leave-one-year-out holiday adjustment.
Upcoming Canadian holidays
So what: the holidays ahead, by brand and province, each with the range history gives it.
Notes
The full forward holiday calendar over the trustee horizon — dated, by brand and province, each with its historical range and whether it clears as a governed input. A long reference list, kept at the bottom; scroll within the window.
Date
Brand
Province
Holiday
Historical range
Descriptive context
The shape of the week
So what: how much of revenue the day of week itself sets; the same grouping the forecast uses for each day's expected level.
Notes
Pooled across every available fiscal year — the same day-of-week grouping the baseline method itself already uses to set each day's fair reference. Same day of week does not mean the same expected revenue: this is structural context, separate from the holiday test above and not a tested possibility.
Day of week × fiscal period — avg net sales per store-day
Notes
Darker cells are higher-revenue combinations. Hover a cell for the exact figure and sample size.
Day-of-week curve by season
So what: how the shape of the week changes with the season, all on one scale.
Notes
One small panel per season, all drawn on the same vertical scale — so you can see how the weekday shape shifts across the year in one view, and which season runs biggest, without holding four overlaid lines in your head.
Local events
So what: what nearby events (games, concerts) add to a normal day, measured on past event days and kept only where the pattern repeated on the test dates.
Notes
Measure past event days, test the pattern, project future revenue and turn the result into a local plan. Additional context, not one of the two hypotheses tested against the unexplained gap — its own separate past-year test, same method as Weather. The upcoming dated opportunities these patterns project are shown below.
Past event days tested0Within the local catchment
Historical event uplift—Versus similar non-event days
Whether the effect repeated
So what: an uplift counts only when its whisker clears no effect and the pattern holds on the test dates.
Notes
Each dot is a brand × event-type pattern — its median incremental revenue on event days within the catchment; the whisker is its 90% range. Repeatability is tested at the brand level (one restaurant rarely has enough event-days), so the event-day counts below are brand-catchment totals, not the restaurant's own count in the KPIs above. A whisker that still crosses no effect means the uplift isn't reliably positive; a pattern must also hold on dates kept for testing to be trusted.
What this table tells you
Loading…
Brand · event type
Catchment
Brand event-days
Median incremental revenue
Median uplift
Guests / spend-per-guest
Data that would sharpen event forecasting
Data that would sharpen event forecasting
So what: why most upcoming events read "more history needed", and what would fix it. Only repeating sports have enough history to price today; the one-off concerts and shows that make up most of the calendar can't be measured by a method that needs the same event to recur.
Notes
Not yet received. On past event-days, uplift is driven mainly by distance to the venue and crowd size — both knowable for a future event once it carries a magnitude — so these inputs would move one-off events from context-only toward a directional, then validated, estimate. Method direction is prototyped in the engine (venue-capacity + distance/magnitude model).
The five data asks for event forecasting are listed once, with every other request, on Phase 2 data requests.
Weather
So what: every weather pattern tested across the portfolio, and what a wet or snowy day takes from each meal period; the weather lens on one restaurant-period is on Restaurant drivers.
Notes
Every weather pattern tested across the portfolio, at brand level, with the past-year test that decides whether it is real. The restaurant-period fingerprint and the weather investigation queue are on Store Breakdown → Restaurant drivers.
What this table tells you
Loading weather evidence…
What repeated history says
Notes
Dots show the typical revenue change on these weather days; lines show the historical range. Green sits above a normal day, red below. Whether a pattern cleared the past-year test is shown in the table below — not by colour.
What a wet or snowy day costs — by meal period and venue
Loading…
Notes
Each row is the typical revenue change on such days against the same restaurant's same-weekday typical-weather days within eight weeks, by meal period and venue (dining & bar, patio, off-premise); holiday and event days are excluded on both sides. The status says whether the pattern repeated on the test year. Nothing here enters the forward number: the daily pilot excludes realised weather, and a forecast-weather feed is the precondition for a forward feature.
Meal period
Venue
Typical change
Per restaurant-day
Days
Status
Where growth came from
Royalty Pool basis
Scaddabush leads; Jack Astor’s is flat
So what: scaddabush combines 17.0% all-restaurant growth with 7.9% same-restaurant growth, while the largest brand is unchanged.
Notes
All-restaurant growth (openings and closures included) versus same-restaurant growth, by brand, for April 2026. Same-restaurant growth is the durable signal a distribution decision should lean on.
Brand growth
So what: growth is concentrated in a few brands; the gap between a bar and its marker is what openings added.
Notes
Bars show all-restaurant YoY; the marker shows same-restaurant YoY. A tall bar with a short marker means the growth came mostly from opening restaurants.
All-restaurant YoYSame-restaurant YoY (marker)
Growth leans on the small brands
Blended same-store growth of +2.8% is really a split: Scaddabush is compounding at +7.9% same-restaurant on real momentum — not just new openings — while the flagship Jack Astor's ($14.4M across 36 restaurants) is flat (−0.1%). So the durable rate currently rides on the smaller brands continuing to scale; Jack Astor's is the single biggest lever on the portfolio total, and the one to watch for whether that durable rate holds or turns.
Ahead of the market on total, behind it same-store
Counting all restaurants, SIR's network revenue grew +5.3% — above the ~+4.3% nominal growth in Statistics Canada restaurant sales for SIR's provinces (latest tax-holiday-clean month). But that lead is openings: SIR's comparable-restaurant growth (+2.8%) sits below the market's nominal rate, so like-for-like SIR is trailing the category even as new builds push the total ahead.
By brand, month by month
Notes
Each cell is the year-on-year revenue movement for one brand and month — the quickest scan of where the portfolio is moving. Hover for the dollar bridge; narrow the window with the period filter.
Below same month last yearNear flatAbove same month last yearEach cell = headline revenue YoY for that brand-month
So what: the rate to plan on is the comparable growth; openings and closures are added on top of it.
Notes
The headline all-restaurant number blends three different things: comparable growth from restaurants open in both years (the real health signal), revenue from new openings, and revenue no longer earned at closures. This separates them for the trailing 12 complete months, all brands — the one read that says whether the royalty pool is growing on its own or on capital spend.
Where the year-on-year change came from
So what: the bar splits the year-on-year change into existing restaurants, openings and closures; the existing-restaurant piece is the rate to plan on, the rest is built.
Organic vs headline, month by month
Notes
Two year-on-year lines for all restaurants, month by month: the headline number everyone quotes, and the comparable (same-restaurant) line beneath it. The shaded space between them is growth bought by opening and closing restaurants — watch whether that gap widens and whether the comparable line holds above zero.
Where to dig next
Royalty Pool basis
So what: the few restaurant-periods still carrying unexplained revenue are the shortlist to investigate (filtered by brand below), and the weekday and meal-period map says where the revenue, guests and spend per guest sit.
Where to dig next — the still-open revenue
So what: most restaurant-periods are explained; the few still carrying unexplained dollars are ranked here, largest first, and are the ones to look into.
Notes
Most restaurant-periods are already fully explained by the evidence tests; only a handful still carry revenue movement with no accepted explanation. Those few are ranked here by the dollars still open — the shortlist worth investigating. Bar length is the unexplained amount (CAD) for that one restaurant-period; colour is its largest traced driver.
Guest trafficSpend per guestInternal patternExternal patternUnattributed
Governed viewResidual reanalysis output
Day & time opportunity map
So what: the day and time map shows where the revenue, guest and spend-per-guest pool sits by weekday and meal period, so the cell with the largest open pool is where a change is worth most; it activates when day-part data is supplied.
Notes
Where the revenue, guest and spend-per-guest pool sits by weekday and meal period — the base for staffing, capacity and daypart promotions. Activates automatically when daypart transaction data is supplied.
Roadmap
So what: what this proof of concept already shows, and what a subscription would unlock next; built on what has been tested.
Notes
The go/no-go picture: what this POC already proves, and what a subscription would need to unlock next. Built on what Analytics has validated so far — not aspiration.
Commercialisation path
Prove the decision before scaling the model
Today: explain historical performance. Next: test forward decisions. Scale only after the challenger improves on the baseline on test periods.
—maximum typical weekly error for a promotable challenger relative to the current baseline
Forward scenarios, local measurement and a challenger forecast used alongside the baseline—not instead of it.
Commercial Phase 2Monitored decision product
Forecasting, daypart operations and margin-aware actions with retraining, drift controls and human approval.
The maturity ladder — current confidence to what a subscription unlocks
So what: the ladder says what is usable today and what a subscription adds; nothing moves up a rung without a past-year test.
Notes
One progression, not two separate lists: what's blocked, what needs a human today, what's fully governed and usable now, and what Phase 2 investment adds beyond that.
On request — conversational Q&A ("ask a question")
So what: a question asked in words would be answered from the governed figures, never a new estimate; a Phase 2 build, not live today.
Ask-a-question — available to build on subscription
A conversational interface that answers trustee questions in plain language against this cockpit's own governed numbers — for example "why did revenue change this period?", "how much headroom is there on the senior covenant?", "which brands are below plan?" — routing each answer to the governed page and the evidence behind it, and returning "not answerable yet" rather than generating a figure. The retrieval and routing are prototyped here; the full interface is built out on subscription.
Forecast · Pool level
Royalty pool · roll-ins dated
The royalty pool’s next 52 weeks, the band each week has earned and the track record behind it. How to read every figure on this page: Notes and Requirements → How this is calculated → The forecast.
Notes
The forward revenue projection for the royalty-pool restaurants. It is built bottom-up — a prediction for each restaurant on each future date, summed across the royalty pool and extended along SIR's fiscal calendar — using the same governed method proven elsewhere in this cockpit. A measured local-event uplift is included where the evidence supports it; holiday and weather effects are held out of the number, not baked in. Each week carries a coverage-validated prediction band — a split-conformal range calibrated on the test weeks of the back-test, whose realised coverage is checked and reported in the readout. How accurate the method has been — and what it would take to replace it with a machine-learning model — is documented under Notes and Requirements → How this is calculated.
The forecast, going forward
Notes
The governed method whose track record is shown below, extended forward along InvestMint's derived reading of SIR's fiscal calendar. The dashed grey lines show the prediction band — a split-conformal range calibrated on the test weeks of the back-test, with its realised coverage reported in the readout below.
ActualForecastPrediction band (dashed)
Management’s budget against the cockpit’s forecast
What moves the next 52 weeks: each layer of the forward line in dollars
How accurate the forecast has been
Notes
What the governed method would have predicted each week, laid against what landed, on a past year kept for testing.
ActualForecast method
Held out of the forecast: measured holiday and weather effects
Notes
Holiday and weather associations are estimated on history and checked with a past-year test, but they are not added to the forward forecast (only a measured local-event uplift is). Bars show the typical measured revenue effect versus a normal day; the past-year test decides whether each could become a governed input. Effects that have not cleared stay context only.
Holiday effects, typical revenue vs a normal day
Weather effects, typical revenue vs a normal day
Forecast · Restaurant level
All restaurants · pool membership shown
So what: each restaurant’s forward year is the all-restaurant line split by fiscal period; a restaurant well off its own history is the one to open, and its royalty-pool status says whether its revenue reaches the royalty. The pool total and the royalty are on the Pool level page.
Notes
The restaurant lines are a share of the all-restaurant forecast line on the Pool level page — each restaurant's share of the reconciled all-restaurant total for each period: the daily pilot's share when the pilot is promoted, the governed method's own prediction for the restaurant (its forward store-days summed by week and scaled to the line) otherwise — so the restaurant figures add back to that total exactly (restaurants outside the pool and openings awaiting their January adjustment are included, and marked); the measured local-event uplift on the Pool level line is carried at pool level and not allocated to restaurants, and the reconciliation between the two pages is stated above the table. On the pilot's split each restaurant's weekly line carries the pilot's 80% band (its daily P10 and P90 summed within the week, then scaled with the week's period to the reconciled restaurant-period forecast; summed daily bands are wider than the exact weekly band unless the days move together, so the band is conservative); on the governed split there is no restaurant-level band, the 80% band being calibrated on the pool line only. Store-level numbers are for review, not distribution. History is each restaurant's actual POS net sales for complete fiscal periods (the guest basis's completeness flag); a period in progress is not shown as an actual. The weather ahead is context only and does not enter the forecast.
Restaurant forward forecast — by week, with the 80% band
Notes
Actual (POS net sales, by week)Forward forecast80% band (P10 to P90)
Show as table
Week beginning (Monday)
Fiscal period
Forecast
Low (P10)
High (P90)
Every restaurant — the forward year
Notes
Each restaurant's forecast summed over the first thirteen forward periods, its royalty-pool status, its share of the all-restaurant total over the same periods, and its actual POS net sales for the same labelled periods a year earlier when all thirteen are complete. Sorted by the forward total; the table downloads as CSV.
Show as table
Restaurant
Brand
Forward 13 periods
Royalty pool
Share of all-restaurant total
Average per period
Same periods, prior year
Forward vs prior year
Weather ahead — the next two weeks by exposure Context only
Loading…
Notes
Each day is the provider's forecast at every restaurant, ranked within the provider's own past forecasts for the same region and time of year (the same day of year within 15 days, across the previous-runs archive's years, so forecast values are ranked against forecast values and a late-month day against its own season) and labelled with the same rule the Weather tab uses, so a snow day ahead means what a snow day behind means. The priced effect is what such a day has typically cost per restaurant by meal period on the test-year lens, times the restaurants exposed; it is not added to the forward number. Nothing here enters the daily pilot until the feature tournament's forecast-backed weather family clears.
Day
Weather
Restaurants exposed
Where
Typical meal-period effect
Indicative revenue
How this is calculated
So what: every figure on the other pages traces to a tested method; this is where to check how one was made before it is relied on. The six drivers, the accounting bridge, and the line between SIR facts and InvestMint estimates.
Notes
See the six business drivers, the exact accounting bridge and the line between SIR facts and InvestMint analysis — the rigor behind every "so what" elsewhere in this cockpit, so you can tell a supported number from an unsupported one before relying on it.
Method in one sentence
Start with SIR facts. Set a fair reference. Keep only explanations that pass the test.
So what: every count is right on its own basis and the bases differ, so a count is quoted with its basis or not at all.
Notes
The board package counts the whole SIR estate; the governed revenue analysis counts Royalty Pool restaurants with revenue in the latest complete month; comparable (same-store) figures count only restaurants open in both years; the day-part map counts pool restaurants with sales in its trailing window; the restaurant master lists every restaurant SIR has ever reported, including closed ones. Quote the basis with the count.
Basis
Count
Where it is used
The forecast, how to read it
So what: the forward line, its band and the layers on top in the order they are applied; read this before quoting a forecast figure.
SIR owns
Official client facts
InvestMint builds
Reconstruction and evidence
From fact to decision
So what: each step answers one question and a figure skips none of them; a claim without its step is not in the cockpit.
Each step answers one question. The same revenue movement cannot be used twice.
How we test a possibility
So what: a possibility reaches a decision only after a past-year test; everything else stays context or investigation.
Every evidence-tested page in this cockpit — Weather, Calendar, Events, the two pattern tests on The gap, the Openings & closures ramp curve, the ML challenger gate — follows the same four-step logic, not a different method each time. Analysis only goes deep where the logic is uncertain, never broad for its own sake.
1 · PossibilityA specific explanation for part of the gap that can be tested and ruled out — not "let's see what the data shows."
→
2 · What would have to be trueThe condition that must hold for the possibility to be sound, stated before looking at the result.
→
3 · What we checkedThe one test that checks that condition — a past year, past dates or a left-out restaurant kept out of the build.
→
4 · ResultCleared or not. A possibility that does not hold stays excluded, however plausible it looked going in.
This is why some pages test a relentlessly narrow question instead of exploring everything available — an untested possibility, however interesting, doesn't get to influence a number you rely on.
How the gap is analyzed
So what: the five steps that split a revenue gap into what is explained and what is still open.
The five steps behind The gap — from the fair reference to what stays unexplained.
1 · Fair referenceExpected revenue sets the starting point.
→
2 · Observed gapActual minus expected. Not yet a cause.
→
3 · Evidence gatheredPrice and guest data assembled on Price & guests, feeding the possibility tested next.
→
4 · What we checkedEach possibility must reduce error on test periods — same test used on Weather and Calendar.
→
5 · Open questionUnsupported revenue stays unexplained — flag it before relying on it for forward planning.
How to read evidence status
So what: a status word says how far a number can be relied on; decision-ready is the only one a decision rests on.
Every number and finding in this cockpit falls into one of five tiers. The Evidence Matrix (Notes and Requirements) uses these directly as its status column; the same colors are used everywhere else.
Reported Fact
Directly reported by SIR — actual revenue, actual guest count. Not calculated, not tested.
Accounting Explanation
Reconciled from reported data by identity, not tested on a past year — the price/traffic bridge, the openings-and-closures effect. True by construction, not evidence.
Evidence-Supported Insight
A pattern that cleared the past-year, past-dates or left-out-restaurant test — real, but it shows what moves together, not proof of cause.
Decision-Ready Finding
Evidence-supported and strong enough that it's been applied to a live number you rely on — today, that's the local-event forecast adjustment.
Investigation Only
A pattern that hasn't cleared the check yet, or hasn't been tested at all. Real information, but don't build a decision on it.
The six drivers
So what: the six business drivers every page reads through; pick one to see its state and where it is measured.
Select a driver. See the business question, calculation, source and limit.
Full methodology, for reference: core equations, guest counts, the budget in the forecast, the daily forecast, accuracy and the model gate, the data wish list, weather regimes
Core equations
The split adds back exactly to revenue change. It does not prove cause.
How SIR GuestCount is used
SIR GuestCount is the official traffic input. Its definition can be confirmed or replaced without changing the method.
How we use management's budget in the forecast
A note of approach, not a requirement. Management's bottom-up budget is measured alongside InvestMint's baseline on the same financial-statement revenue basis, and we use it as an input: we keep their pool total (their strength) and only decide how it splits across restaurants, rather than trying to out-forecast it. Figures are computed live from this run's plan-vs-actual data.
Bottoms-up daily forecast controlled pilot
Can we forecast each restaurant, each day, and still have it add up to the pool total? This is the first test — checked against periods kept out of the build, with a low-to-high range on each figure. Shown only after the daily forecast has been run for this run.
How accurate the forecast is, and what a machine-learning model would need to improve on it
The forward forecast on the Forecast tab comes from the governed baseline method. These are its measured accuracy figures and the bar any future machine-learning model would have to clear before it could replace the baseline in front of trustees. No machine-learning model is in production today — the governed method is what's live.
Typical weekly error—Current historical baseline
Share of periods we could check—test periods, kept out of the build
Forecast weeks available—Fiscal periods beyond the last actuals
Machine-learning model—no production forecast until proven
Phase 2 data wish list
Only inputs that change a decision are requested. The ranking is generated from the run's capability gaps.
Priority
Data requested
Required grain
Decision unlocked
Client effort
Future machine-learning model — specification
Model family
Operating boundary: until every check passes, any machine-learning model stays experimental and the governed method remains the reference.
Weather regimes — how a day is classified
“Warmer than typical” and “colder than typical” are relative regimes, not absolute extremes. Each store-day’s mean temperature is converted to a robust z-score against the distribution for that province and time of year (the same day of year within 15 days, across the years); a warmer/colder score starts at |z| > 0.75 and saturates at |z| = 3.0. Heavy rain needs ≥ 2 mm and a precipitation percentile ≥ 0.65 within the same province-month; heavy snowfall needs ≥ 1 cm and a snowfall percentile ≥ 0.65; high wind needs a gust ≥ 40 km/h and a gust percentile ≥ 0.70. The day’s primary driver is the highest of these scores when it reaches 0.25 — about |z| ≥ 1.3 for temperature — otherwise the day is “typical.” Weather source: Open-Meteo ERA5 daily history at the province centroid.Snowfall status: the snowfall score is computed, but the current weather import carries no snowfall observations, so no heavy-snowfall days are flagged; once an import with snowfall populated lands, they appear as their own driver automatically.
Data Glossary
So what: one name per number; when two pages seem to disagree, check here whether they name the same thing. Every metric with a plain-language definition, and the labels that mean the same thing.
Notes
One name per number. Every metric used across this cockpit with a plain-language definition — and, where the same quantity has appeared under more than one label, the words that mean the same thing. Financial and operational metrics first; the method and evidence terms follow.
Naming rules used everywhere in this cockpit: Revenue is the reported top line; a change is always a period-over-period delta, never a level; comparable / same-restaurant excludes openings and closures; spend per guest is the standard term for average check; and pooled means statistically aggregated, not SIR's royalty Pooled Revenue.
Phase 2 data requests
So what: each request unlocks an analysis the current data cannot support; the ones that would change a decision are the ones to raise first.
Notes
Data InvestMint would request from SIR to sharpen the Phase 2 proof-of-concept. Each item would unlock an analysis the current governed data can't yet support — not yet received.
Restaurant populations — one estate, several filters
So what: every count on a page names the population it counts; two figures that differ by a few restaurants usually differ by filter, not by data.
Every count the pages use, and the restaurants that make each one differ from the royalty pool. “All restaurants” on a pooled page means every restaurant in the pool that month; on the operating basis it means every restaurant trading.
Population
Count
Who is in it
Royalty pool, 2026-08
52
the base every other population is read against: restaurants in the pool in that month, roll-ins dated
Trading in 2026-08, operating basis (all brands, pool or not)
57
the pool plus, outside the pool: Abbey's Minett, Duke's St. Lawrence; trading before their January roll-in: Jack's Oshawa (joins 2027-01); Scaddabush Aurora (joins 2027-01); Scaddabush Windsor (joins 2027-01)
Restaurant-level forecast (every restaurant the daily panel carries)
57
the same restaurants as trading
Roster locations with coordinates (closed ones included)
62
trading plus the closed restaurants still on the roster: Jack's Greenfield Park, Jack's North York (Yonge St), Reds Fallsview, Reds Wine Tavern, Scaddabush Mimico; 1 roster row(s) without coordinates
Weather feed (realised and forward)
61
the roster locations less Scaddabush Aurora (coordinates joined after the last weekly weather extension; in from the next one)
Flagged pool-eligible on the roster
58
the pool plus the closed restaurants that were pooled: Jack's Greenfield Park, Jack's North York (Yonge St), Reds Fallsview, Reds Wine Tavern, Scaddabush Mimico; flagged ahead of their roll-in: Jack's Oshawa
With meal-period data (day-part reads)
53
trading less Abbey's Minett, Duke's St. Lawrence, Scaddabush Aurora, Scaddabush Windsor
Package portfolio (Nov 10, 2026)
57
SIR's own count of its restaurants in the Board package, on the package's date and definition; shown as reported, not derived here — the operating data above is the cockpit's population
So what: where a guest count changed definition, guests are modeled from revenue until SIR confirms the counting rule, so a guest read there is an assumption; revenue is unaffected.
Notes
A data-definition question surfaced by the guest-basis check — a clarification of existing data, not a new dataset. Guest-based reads use a modeled guest count in the interim; open until SIR confirms. The break test and the modeling rule are on the Methodology page (Method notes).
What we need from SIR
Why it matters & how it’s handled now
Confirm what changed in Red’s guest counting effective FY2026 — e.g. a POS or metric change, bar covers now counted, or a store format reclassification. With that confirmation the modeled guest count below is replaced by the corrected actuals.
At Red’s single comparable store, FY2026 reported guests step up ~+40% and spend per guest ~−25% every period while revenue is continuous (+5.2%) — the signature of a guest-count definition change, not real behaviour (baseline still under review: OFFICIAL_SIR_BASELINE_DEFINITION_PENDING). Revenue and the royalty-pool reconciliation are unaffected. Interim treatment: rather than drop Red’s, its guests are modeled from revenue — hold the cheque on its pre-break definition (each FY2026 period’s prior-year same-period cheque, preserving seasonality), grow it by the portfolio’s real price signal (the clean brands’ average cheque growth), and divide the trusted revenue by that cheque. This re-includes Red’s at a consistent split: modeled guest traffic is roughly flat (≈ −0.2% YoY) versus the reported +40%. Assumption: Red’s real cheque grew in line with the portfolio; any reasonable price assumption still lands guests near flat, nowhere near +40%, so the conclusion (no real traffic surge) is robust while the exact figure depends on the anchor. Reported figures are retained in the data for audit.
Method notes, for reference: how a few reads are treated
Method notes — how a few reads are treated
So what: the data-treatment choices behind the tabs, gathered once; a figure that looks off is checked against these before it is questioned.
Notes
Data-treatment choices that sit behind the analytical tabs, kept here rather than repeated on each tab.
Read
Treatment
Period comparisons after a 53-week year (every period read)
Menu price vs dining-out inflation (Inflation & real growth)
The dining-out CPI comparison excludes the GST/HST-holiday months (Dec 2024–Feb 2025), when restaurant meals were temporarily tax-free and the published dining-out CPI is tax-distorted. Excluding them keeps the menu-vs-inflation gap on a like-for-like basis.
Guest basis (every guest-based read)
Prior-year comparisons match period days: each traded day is compared with the same period-day a year earlier, so a period in progress is compared with the same number of days and the 53-week FY2025 P13 is compared week for week. A brand’s guest count is modeled from revenue from the first period where spend per guest moves more than 15% or guests more than 25% year on year and the move persists (at least two periods, four in five thereafter): the cheque is held at the last pre-break same-period cheque, grown by the clean brands’ realised price signal, and revenue is divided by it. Both sides of a comparison are modeled when both fall after the onset. Reported figures stay in the data for audit.
Store-level traffic vs price (Traffic vs price)
A restaurant-period closed more than two days below its normal operating pattern (renovation/closure) is set aside for that period, so a temporary shutdown isn’t read as a demand change.
Channel blend in the traffic ÷ price split (Traffic vs price, Price & guests)
The decomposition uses one blended spend per guest — total comparable revenue ÷ total SIR GuestCount — so it does not read the channel mix, and splitting off-premise into take-out and delivery does not change any traffic/price figure. But delivery revenue is in that revenue total while a delivery order is not a dine-in cover, and it is unclear whether SIR’s guest count includes delivery orders at all (SIR hasn’t yet confirmed how guests are counted). So when more sales come through delivery, spend per guest can edge up on its own — a change in the mix of channels looking like a menu-price move. Read the price effect alongside the delivery share on Price & guests; the Price & guests readout flags periods where the two move together. A channel-clean split needs channel-level guest counts (see the data request below).
New-restaurant maturation ramp
The early-life age→revenue ramp is fit as a small candidate tournament — a parsimonious two-parameter saturating ramp against a coarse-binned monotone curve — and each is scored by a left-out-restaurant test (each restaurant scored by a curve built without it) against assuming the new restaurant is already mature. The leading candidate is promoted to a governed figure only if it clears that gate; its current status is shown on the Evidence Matrix. The winning curve is the basis for the Scenario planner’s opening/closure what-if.
Renovation effect
SIR has no standalone renovation dataset, so a renovation is inferred from the daily operating signal — a run of consecutive closed days for a restaurant (the operating-day flag already reflects the curated “0 Sales Days” closure log) — and classified by closure length: a short closure (3–6 closed days, cause unknown — health/equipment/staffing/weather or a single-store data gap, usually not a renovation), refresh (7–10), major renovation (11–20), re-brand (over 20 days with a concept change) and extended major (over 20 days, no concept change); a closure under 3 days is excluded. Only the four renovation classes feed the recovery / matched-control read. A estate-wide-outage guard sets aside any run that falls mostly on days when more than 30% of the operating restaurants are simultaneously dark — a portfolio-wide data-coverage gap, not a real closure (a renovation closes one restaurant, not the whole estate). The per-event register (Store Breakdown → Store Renovations) carries each event’s class, concept before/after, and — where a curated closure record covers the run — its reason, marking the row as from the closure log rather than inferred from a gap. Revenue after reopening is measured against that restaurant’s own pre-closure baseline and scored by a left-out-renovation test (each renovation scored by a curve built without it), on the same tournament + gate as the maturation curve. On the current data the fit has no time-trend — it is a level shift (post-reopening revenue settles a few points above the pre-closure level and stays there), not a recovery ramp. Clean baseline: the pre-closure baseline is the median operating-day revenue in a 120-day window that ends four weeks before closing, so a wind-down slowdown right before a planned closure cannot depress it and manufacture a false uplift. (Because it is a median, a short wind-down barely moves it anyway; the exclusion is a robustness guard.) Matched-comparison read, in three statements.Method: each renovated restaurant's pre→post change is compared with the nearest same-brand, similar-size restaurants over the same calendar months. Result: the difference is the association reported on Renovations, with a bootstrap 90% band. Limitation: the matching takes calendar, seasonality, the estate-wide trend and the restaurant's own level out of the comparison; it narrows the comparison and does not establish what the restaurant would have earned without the work, nor the return on it. Its governed status — and, when that matched-control effect is significant, its decaying application to the forward forecast — is shown on the Evidence Matrix, Store Renovations and the forecast readout.
Is discounting working? (Price decomposition)
The estate-wide discount read demeans each brand's discount rate and its revenue / guest / spend-per-guest growth within that brand (brand fixed-effects), so the correlation measures a brand discounting more than its own norm, not brands that sit at different baseline discount levels. It is a governed discount figure (FS_Actual Promo Discounts ÷ gross revenue) joined to the governed driver attribution, channel mix and matched local-event days. The panel is associational and investigation-only — a modest cross-section (≈38 brand-periods across the measurable brands, a few fiscal years), no causal claim; discounting could be a symptom of a softer period rather than its cause. The channel and event reads are the same within-brand correlation on in-restaurant share and local-event exposure.
SIR vs industry — co-movement / correlation
The correlation of SIR's growth with the industry benchmark is computed on raw year-on-year over the months clear of the GST/HST tax-holiday base effect (the same months excluded from the trend line and the relative-performance bars). It is a small sample — directional context, not a significance test: SIR's year-on-year begins in late-2024, so its overlap with the benchmark is ~10 clean months. Alternatives considered and not adopted: (1) a 2-year annualized basis (each month compounded with the same month a year earlier), which would cancel the tax-holiday base on both series and keep the Dec 2025–Mar 2026 months — but on this short overlap it collapses to ~3 usable points, too few to correlate; (2) including the tax months on raw year-on-year, or (3) a mixed basis (normalize only the tax months) — both let three to four distorted points dominate a ~13-point sample and pull the correlation toward zero, so neither is a faithful read. The tax months are shown greyed on the scatter for transparency but kept out of the fit.
Data that would sharpen this view
So what: requested from SIR; each would let a context-only read become a tested one.
Requested from SIR — would let dollar mix become a per-seat utilization view.
Data
Use
Patio seating capacity (seats, open months)
Would unlock per-seat patio utilization, not just dollar mix — same pattern as the Openings & closures real-estate data request
Data that would sharpen event forecasting and the scenarios
So what: with these, most upcoming events could be priced rather than listed.
Requested from SIR — the event-forecasting asks (Demand & calendar → Events) and the scenario-planner ask, gathered here so each is listed once.
Data
Use in analysis
Venue → seating-capacity reference table
Gives every upcoming event a crowd-size magnitude (the venue is already known for 100% of events) — enough to price one-off concerts and shows with a distance × magnitude model, not just repeating sports
Historical concert & performing-arts event-days with revenue
The durable fix: today the entire local-event training history is sports, so non-sports uplift has to be transferred rather than measured
So what: not in any board package yet; it would set SIR’s growth against the market on one basis.
Not in any board package; needs an external feed. Requested following the September review.
Data
Use in analysis
TSX daily unit price and trading volume history for the Fund’s units
Daily trading volume and distribution yield over time on the Distributions subtab, and a properly dated unit-price reference (the undated June 8 price tile was removed)
Next actions
So what: what is ready to act on and what needs review first; acting and measuring becomes next period's reported numbers.
Notes
Iterate: see what is ready for use and what needs review before it becomes a decision. Acting on one of these and measuring the result becomes next period's entry on the Reported revenue change tabs — the cycle repeats, it doesn't end here. Each question below is also answerable directly via the Ask a question button in the header.
Review queue
So what: the items to confirm before relying on them; the backlog behind Next actions.
Notes
Items that need confirmation before you can rely on them — restaurant-periods worth a first look, and the operational backlog behind Next actions.
POS ↔ FS reconciliation — data quality
So what: within 0.5% is a clean reconciliation; a larger gap is where the two revenue bases diverge and the first place to look when a figure is challenged.
Notes
SIR's governed check that point-of-sale GrossSales agrees with financial-statement Gross Revenue, per restaurant and fiscal year. Within 0.5% is a clean reconciliation; a larger gap flags where the two revenue bases diverge. This is a trust read on the inputs, not a business result.
Mean gap by fiscal year
So what: a year whose mean gap sits above the tolerance is one where POS and FS do not agree; the rows beside name where.
Where to look first
So what: the first places to look when a figure is challenged, in the order that resolves most.
Actual revenue, expected revenue, known activity and weather remain separate.
Restaurant
Period
Observed revenue (CAD)
Expected revenue (CAD)
Documented activity
Weather context
Review queue
So what: the open items by priority; a blocking item is what holds the trustee-ready status.
Priority
Item
Context
Required response
Evidence Matrix
So what: every possibility the cockpit has tested, its status, and what it is used for.
Notes
Every possibility this cockpit has tested or investigated, its current status, and what it's used for — see How this is calculated for what each tier means.
Decision-ready—Applied to a live number
Evidence-supported—Cleared a past-year test
Investigation only—Not yet cleared, or not tested
Possibilities tracked—Across the whole cockpit
Every possibility, one status
So what: live rows read fresh from this run; the last two are a marked point-in-time investigation, not a live read.
Notes
Live rows are read fresh from this run's real data. The last two rows are a point-in-time investigation, not recomputed every run — clearly marked.
Possibility
Tier
Evidence this run
Estimated impact
Where to see it
Management use
Data that would sharpen this further
So what: every request already flagged across the cockpit, gathered once; the ones that would move a possibility from context to decision-ready come first.
Every specific data request already flagged across this cockpit, gathered in one place. None of it blocks what's shown today — each row would extend a real, already-working view. See Roadmap for the full model-ranked wish list.
Data requested
Would unlock
Where it's used
Square footage, seating capacity, location type, lease dates, remodel history
Revenue-per-square-foot productivity, capacity ceiling, and forward closure risk
Openings & closures
Patio seating capacity (seats, open months)
Per-seat patio utilization, not just dollar mix
Price & guests
Price-change event log (effective dates of actual menu price changes, ideally per item/store)
A real, validated price-elasticity model in place of the current descriptive view
Scenario planner
Daypart transaction grain (revenue, guests, check by meal period)
The full day-and-time opportunity map, not the design preview
Explore
Channel-level guest counts (dine-in / take-out / delivery, per store-period)
A channel-clean traffic-vs-price split — isolates a delivery-mix shift from real menu pricing and puts spend per guest on a consistent cover basis
Price & guests
Scenario planner
So what: a what-if tool for a meeting; every number is looked up or projected from this run’s own history, with the same status labels used across the cockpit.
Notes
A facilitated what-if tool for live discussion. Every number below is looked up or projected from this run's own historical data — never generated — and carries the same status labels used everywhere else in this cockpit. Complements Roadmap's longer-range forecasting path with something usable today.
Event impact at a restaurant
So what: what a local event of this kind has added here before, with its status: Passed the past-year test, Context only, or Not cleared.
Notes
Possibility: this event boosts revenue at this restaurant. What would have to be true: the combination selected would need to clear the past-year test on Events to be a governed input. Combinations that have matched history but have not cleared that check are shown as Context only / Not cleared — the observed pattern is displayed for local planning, but it is not a governed input and does not feed the covenant roll-up unless you include it deliberately. Only when no matched event-day exists for the combination is nothing shown (see Calendar & events for the full picture and the upcoming dated opportunities).
Typical uplift—Median effect on similar days
Incremental revenue—90% range
Guest uplift—Change in guest count vs a normal day
Status—Past-year test result
Open or close restaurants
So what: the net dollars an opening or a closure would add or remove, projected from the ramp new restaurants follow; illustrative until the ramp clears its past-year test.
Notes
Possibility: opening or closing N restaurants nets a predictable dollar effect. What would have to be true: the maturation curve's own left-out-restaurant test (see Notes & Requirements) would need to clear its bar.
Typical new-restaurant revenue—Once mature, per period
Ramp at selected age—Share of mature run-rate
Typical closure effect—Per closed restaurant
Projected net effect—Openings minus closures, this period
Change price by X%
So what: what similar price moves did here before, as a median; not a prediction until a tested price-response model lands.
Notes
Possibility: a price move of this size changes revenue predictably. What would have to be true: a validated price-elasticity model (requested in Notes & Requirements) (see Price & guests for the underlying driver data).
Current spend per guest—Most recent period
Similar historical moves—Periods found
Historical revenue range—What happened those periods
Median revenue move—Not a prediction
Unusual weather this period
So what: what revenue did on days with this weather, as history shows it; an association, not a cause.
Notes
Possibility: unusual weather this period shifts revenue by a predictable amount. What would have to be true: an association, not a causal claim. Real historical variance for the selected brand and weather driver only (see Demand & calendar → Weather for the full pattern library).
Historical revenue variance—Interval
Illustrative revenue swing—Applied to recent period revenue
Exposed days—Matched historical days
Status—Past-year test result
Build & compare a scenario — combined roll-up
So what: tick the levers to include and read their combined full-year revenue effect, the royalty at the pool’s 6%, a directional read on distributions, and where the bank covenants stand. Save and compare several scenarios side by side.
Notes
Every dollar is looked up or projected from this run's own history in the panels above — never generated. Structural levers (restaurant count, price) are held across the year; weather and events are applied the number of periods or days you set. Distributions are directional (the Fund distributes ~100% of distributable cash over the year; exact flow-through needs the Fund's cost structure). Both leverage covenants are governed and scenario-responsive — from the Board covenant certificate (P12 FY2026, rolling four quarters; public Q3 F2026 filings corroborate), the scenario's revenue change flows to covenant EBITDA at the selected flow-through and moves senior and Fund leverage against their caps. Fixed-charge coverage shows its governed baseline; its scenario move needs the certificate's Funded Debt Service (see Notes & Requirements).
Net pooled revenue effect—Full year · — per period
So what: the data that would turn these scenarios from illustrative into tested.
Notes
Requested from SIR — would let these move from illustrative to validated.
The price-change event log this needs is listed, with every other request, on Phase 2 data requests.
Observed facts, explanations, discoveries and decisions remain separate.
Diagnostics — internal
So what: a method reaches the cockpit only when it ranks first here on test periods; the tournaments and the business-cost ranking are the evidence, kept off the trustee spine.
Notes
These panels support the Forecast and Openings & closures reads but are too detailed for the main tabs, so they live here in the appendix. Several activate only when their governed data lands; each carries its own status label.
Forecast diagnostics
The forecast's accuracy read (backtest error, coverage, promotable bar) lives on the Forecast tab and is not repeated here. Deeper past-year tests sitting behind the forward projection — whether one champion is the right frame (model tournament), whether reconciling the forecast across store, brand and pool levels helps, what an asymmetric, peak-weighted business cost says about the ranking, and how much of the gap every driver together reduces on test periods, and how much no validated family reduces (validated-feature test). Diagnostic context, kept out of the main Forecast view.
More diagnostics, for reference: the validated-feature test, the site card, same-brand proximity, roll-in economics
Site & Market Intelligence
A location-aware read on each restaurant, built from the governed roster (coordinates, brand, pool status) and the performance data already in this cockpit. It asks not just “did revenue rise?” but whether a restaurant performs in line with where it sits. Spatial links are investigation only — proximity is exposure, not proof — and nothing here changes a governed number. Capacity, catchment, tenant and footfall inputs are not yet supplied; each is shown as a data request rather than guessed.
New-store economics
How new restaurants earn in over their first periods — the governed roll-in read behind the openings line in the revenue bridge.
Growth is real
New package context — — (—):—
In , every dollar of the +5.3% ($1.22M) adds up exactly in the revenue bridge: existing restaurants grew — like-for-like, with the rest from new openings. Some of the underlying business causes remain open — the Revenue bridge's Why it moved shows how much of the like-for-like change known causes explain. The forecasting method holds on the test year, and a store-level pilot sharpens per-restaurant allocation by about 20%.
Notes
The revenue math reconciles exactly and the forecast holds on the test year, so the numbers reconcile; the drivers are named where the evidence supports them, and — of the like-for-like change is still unexplained. The durable read is — same-restaurant growth; new openings add the rest, and a prior-year data fix explains most of the wider cumulative gap. Below, the same trusted decomposition powers a store-level forecast pilot that sharpens per-restaurant allocation by about 20%.
Governance status: an internal, governed diagnostic rather than a finished client product; the store-level forecast is a controlled pilot, not production; weather and holiday effects are context only; and the underlying source-data agreements still need sign-off. About $0.7M — the one-off share of the wider cumulative change, distinct from the ~$0.6M still open on the latest period alone (see The gap) — remains unexplained once openings and the prior-year fix are set aside. See Notes & Requirements for the full governance state.
What's good
Reported revenue splits into its drivers with near-zero error, and the forecast landed within 7.6% a week of what landed on the FY2026 test weeks (the pool as a whole) — the numbers add up, and the drivers are named where the evidence supports them (— of the like-for-like change is still open). A new store-level pilot, riding management's own pool total, cuts store-allocation error by about 20% (restaurant-level error 9.2%→7.4%).
What to watch
The — headline is mostly new openings (—); the durable, repeatable rate is +2.8% same-restaurant this month — the comparable-restaurant rate. And that same-store gain is led by price, not guests: guest traffic is slipping while the cheque rises, so after inflation the real rate is near flat — read it as price-led, not demand-led (see Price & traffic and Inflation & market). A wider cumulative view flags $4.2M worth tracing, most of it already accounted for (see How much we can explain).
Potential risks
Headline growth leans on openings that may not repeat, so the run-rate could soften if the build pipeline slows. Part of the recent trend may also be weather- or holiday-lifted and prove temporary — we are actively reducing that risk by bringing in additional data to isolate those effects and size how much of the gain is durable.
Board brief — — package (—)
So what: Covenants comfortably in compliance. Growth quality and specific data gaps need operational attention. A plain-language read across the cockpit for the board.
Notes
A written brief compiled from the cockpit for the — board package (—). Governed analytical figures are on the cockpit’s same-restaurant basis through Apr 2026; board-package figures (covenants, distribution, — same-store) are as reported in the package. Refreshes with each package.
Full narrative
Portfolio performance & growth quality
Headline vs organic (all restaurants): network revenue +5.3% (+$1.22M), but openings supplied ~$1.07M of it; the durable same-restaurant rate is +2.8% (governed basis, through Apr 2026). Brand divergence:Scaddabush is the engine at +7.9% same-restaurant; flagship Jack Astor’s is flat at −0.1% (same governed basis; the board’s — segmented-sales file covers a different period and store set). Price over traffic: same-restaurant gains are price-led — the cheque rising while guests are fewer. Same-store sales as reported in the package:— P11 FY2027 laps the one-time World Cup lift of P11 FY2026 — a mechanical comp headwind to strip from next year’s base.
Channel: dine-in is winning the trade
Financial stability
Covenants — —:— Distributions:— An intra-year payout above the ~100% annual target is the normal seasonal shape; the year trues up to the target.
Data accuracy & accounting anomalies
Red’s guest-count integrity: reported guests jumped — while per-guest spend dropped — against — revenue at the reporting onset — guests are modeled from revenue until the POS reporting change is confirmed. Prior-year reconciliation: of the $4.2M cumulative unexplained movement, ~$0.9M stems from a corrupted prior-year period (P2 2025) under reconciliation. XTM fraud: ~$200K receivable from the tip-processing fraud; possible Q4 write-off if unrecovered.
Forecasting & predictive analytics
Model accuracy: pool-level weekly forecast error (WAPE) ~7.6% on the FY2026 test weeks — a portfolio measure; the store-level pilot cuts store-allocation error ~20% (store WAPE 9.2%→7.4%) by splitting management’s pool total with the model’s shares. External signals: local events are the only decision-ready external metric adjusting live forecasts; weather and holiday remain descriptive context. ML status: no production model live yet — blocked until it consistently improves on the baseline error threshold.
3How much of the change is explained
Traffic and spend explain the change; only ~$0.7M is one-off
So what: traffic and spend explain $14.4M; new restaurants and one prior-year data fix account for most of the remaining $4.2M.
Notes
Across all 485 restaurant-months reviewed (each restaurant, each month), we started with $18.6M of revenue movement the basic breakdown — existing restaurants, openings, closures — doesn't explain, then tested what drives it. Only exact accounting held up; no pattern model was strong enough to rely on.
Where the $18.6M revenue gap went
So what: traffic and spend explain 77.6% of the change directly; new-restaurant ramp and a prior-year correction account for most of the rest, leaving under 4% one-off.
Notes
The green block is traced by exact accounting (guest traffic × spend per guest). The amber block looked unexplained, but a deep dive traced most of it to two data-shape issues — new restaurants with no prior-year history, and one corrupted prior-year period being reconciled with SIR. Only a small remainder is one-off.
Traced to traffic & spend per guest$14.4M · 77.6%
Needed a deeper look$4.2M · 22.4%
Start$18.6M of movement the basic breakdown (existing + openings − closures) doesn't explain.
→
Traffic & spend$14.4M traced by exact accounting — how many guests, how much each spent.
→
Mostly accounted for$4.2M, now mostly explained: ~61% new restaurants that have no comparison year yet, ~22% a prior-year data fix (in progress with SIR), leaving ~17% (~$0.7M) one-off.
Breaking down the $4.2M that needed a deeper look
New restaurants, no comparison year yet
~$2.6M · 61%
Structural — a new restaurant can't be compared until it has a full prior year open.
Prior-year data-quality fix
~$0.9M · 22%
A corrupted prior-year comparison period (fiscal P2, 2025). Actionable — being reconciled with SIR.
One-off
~$0.7M · 17%
A small remainder no driver explains — one-off.
Breakdown as of the run through 2026-04 (run_34b8a7003aac42c5). These shares are computed from that run and are refreshed monthly, once the latest month closes complete; each refresh is re-checked against the same bar.
Explained by accounting identity (true by construction)New-store ramp + a prior-year fix; only ~$0.7M one-off
4What has been tested, and what passed
Local events are the only decision-ready signal
So what: events can inform a live number directionally; weather, holidays and operating patterns stay context or investigation until they clear a past-year test.
Notes
Every possible explanation is held to the same bar: it must reduce error on a past year, past dates or a left-out restaurant kept out of the build. This is the scorecard — what cleared that bar, what didn't, and what it's allowed to be used for.
Evidence status
So what: one live signal is decision-ready and one forecast is supported in pilot; the remaining candidates are not used for decisions.
Notes
78 possibilities are tracked across the full cockpit; the ones that move a trustee decision are below. Colour = evidence tier.
Possibility tested
Result this run
Tier
Allowed use
Reported revenue decomposition Total = existing + openings + closures
Reconciles with near-zero error across 93 rows
Reported fact
Governed reference for distribution discussion
Price vs. guest traffic split What drove same-restaurant change
Traced 77.6% of the open gap ($14.4M)
Accounting
Diagnostic — true by construction, not tested on a past year
Local events → revenue Sports / concerts near restaurants
$7.7M historical uplift measured. In the forward forecast: — from effects tested on test weeks, — from effects estimated and not yet tested
Decision-ready tested Investigation estimated
Tested effects applied to a live number; estimated effects applied directionally, read as a range
Holidays → revenue Calendar-adjusted expectation
6 validated + 7 discovered calendar patterns. In the forecast, measured but excluded — on its own it makes the forecast less accurate when tested on past periods
Investigation
Calendar context only — not applied to the forecast number
Weather → revenue Cold / heat / rain / snow
6 of 20 patterns historically supported — but no dollar subtraction taken
Investigation
Descriptive context only
Internal operating / sales-channel pattern
Model ran; not in use (0 added explanation on test periods)
Not in use
Not used
External weather/calendar pattern model
Model ran; not in use (0 added explanation)
Not in use
Not used
New-restaurant maturation curve How openings ramp with age
Diagnostic only — not cleared for adjustment
Investigation
Scenario discussion only
Store-level daily forecast (pilot) Each restaurant, each day
Improves on its baseline by ~13% on test data; reconciled to management's total, cuts store-allocation error ~20%
Supported
Controlled pilot — informs store-level views, not a live production number
Weekly ML challenger forecast
Not attempted — no model risk taken yet
Not started
Blocked until it improves on the weekly baseline
Reported fact straight from SIRAccounting true by constructionSupported cleared a past-year testDecision-ready applied to a live numberInvestigation not cleared / not testedNot in use not cleared / not run
5Can it forecast forward yet?
Weekly error is 7.6%; store allocation improves by 20%
So what: the weekly method is usable for discussion, and the daily pilot materially improves how management’s total is distributed across restaurants.
Notes
The governed weekly method was tested against what happened in fiscal 2026 — genuine proof it works. Separately, a finer store-by-store daily model now runs as a controlled pilot and improves on its own baseline (below). A weekly ML challenger to replace the governed method still hasn't been built, so no production model risk has been taken on.
So what: the weekly baseline differed from actual revenue by 7.6% on average, on test periods.
Notes
Weighted error (WAPE) on the FY2026 test periods — lower is better. The dashed line is the bar a future ML challenger must improve on to be promoted.
Baseline weekly error7.6%Typical weekly error for the pool as a whole on the FY2026 test weeks, not restaurant-level
Forecast weeks available—Scenario discussion only
The bar a challenger must clear—A challenger must improve on the baseline by the set margin and pass every check
Weekly ML challenger—The upgrade meant to replace the weekly method — separate from the daily pilot
Foundation is real
The same method that explains history has been shown to predict the test year’s weeks to within 7.6% — independent of any forward projection.
New: a store-level daily pilot
A separate, finer track forecasts each restaurant each day and improves on a naive “repeat last year’s weekday pattern” baseline by about 13% (on the point-of-sale sales basis — a different measure from the 7.6% weekly figure, not directly comparable; measured on the test year). Its most useful form: keep management’s pool total and split it across restaurants using the model’s shares, cutting store-by-store allocation error about 20% (9.2%→7.4%) while still matching management’s pool total. That sharper store view is the usable result today — still a controlled pilot, not production.
Forward use is for discussion
The 80-week outlook is for discussion, not a committed forecast; a committed forecast waits for a challenger that clears every check.
Forward forecast by fiscal period — and how prior forecasts hold up
So what: as each package lands, the actual is read against the forecast that stood before it was known, and the 4-to-16-month horizon is where a difference matters most. The governed weekly method extended forward, rolled up by fiscal period with its prediction band.
Notes
Forecast points are the Forecast tab’s forward outlook (royalty-pool basis, discussion-only), summed to fiscal periods; the band is the sum of the weekly split-conformal bounds, an outer envelope. Each refresh retains its forecast as a vintage (context/forecast_vintages.json) so later actuals can be compared with the forecast in force at the time; the first vintage is recorded on the next refresh, so the comparison fills in over time.
6What would move this forward
Confirm guest counts first
So what: a low-effort guest-count definition unlocks the cleanest split between traffic and spend per guest; it is the first ask of management.
Notes
Only data that changes a decision is requested. The first ask is low-effort and unlocks the most.
1GuestCount definition
Low effort
Confirm how guests are counted (channels, cross-year consistency). Unlocks a clean split between traffic change and average-check change — the biggest single blocker.
2Campaign execution
Medium
What promotions ran, where and when. Unlocks measuring promotion response and separating it from weather and events.
3Patio seating capacity
Medium
Seats and open months by restaurant. Turns dollar mix into a per-seat utilization view and models weather exposure.
4Margin & promotion cost
High
Contribution margin and discount cost. Unlocks ranking actions on profit rather than revenue alone.
5New concepts: the conversion capex and the concept’s POS line
Low effort
What is usable now
So what: historical reporting is usable internally; forward scenarios remain a controlled pilot; production use requires Phase 2 controls.
Notes
One progression: what's usable now, what a controlled pilot adds, and what a full Phase 2 subscription unlocks.
Available nowHistorical read, checked
Reported change, driver & residual investigation, weather/event/holiday evidence — all governed and usable internally today.
Next validation stageControlled operating pilot
Forward scenarios and a challenger forecast used alongside the baseline — with restaurant teams still approving each action.
Commercial Phase 2Monitored decision product
Production forecasting, daypart operations and margin-aware actions with retraining, drift controls and human approval.
Operational backlog: 16 review items open — 10 performance investigations and 6 data-reconciliation batches (15 rated P1). These are the confirmations needed before individual findings can be relied on.