KPI360 — the till and the books, for restaurant groups

The industry's math

Margins are 3–5%. The invisible leaks are 4–8%. Catching them is the whole game.

Running restaurants has never been tighter: 42% of operators weren't profitable in 2025, and total costs are up 36% since 2019. When the leak is bigger than the margin, the difference between a profitable year and a losing one is simply whether anyone catches it. That is KPI360's entire job — watch every order and every shift, flag the drift the week it starts, prescribe the fix, and measure what fixing it was worth.

42%

of operators unprofitable in 2025 — up from 29%

+36%

cost growth for the average restaurant since 2019

4–8%

of revenue lost each year to undetected variances

Sources: National Restaurant Association State of the Industry & 2025 Operations Report; industry cost surveys, 2019–2026. The 4–8% figure is precisely the leak KPI360's watchdogs, drift checks, and morning briefs exist to catch.

Proof · 12 locations live

What it's already caught — on real money.

KPI360 runs in production today across 12 restaurant locations, $20M of annualised sales and 603,000 transactions — groups running full cost sheets alongside groups running nothing but their POS. Every number below came out of that live data, not a model — counted on 24 September 2026, the way the screenshots below are dated.

62.5%

one store's prime cost, flagged 9 points high in week one

+20% YoY

like-for-like sales across a full six-location group, every week

$13k/wk

3P commission exposure surfaced, renegotiation drafts ready

$370k

of missing sales history found and restored during onboarding

24 months

of order history pulled on connection — and growing weekly

$187/hr

sales per labor hour, measured straight from the POS clock

The actual product

Not a mock-up. These are photographs.

Captured from the live demo by a script when a release ships — refreshed at release time, not daily, and never rotated — so what you see here is what ships. Open the demo yourself and click any number on any of these screens. Click a photograph to see it full size.

The Pulse — What is wrong, what it costs, what to do, and whether the last fix worked. — The Pulse
What is wrong, what it costs, what to do, and whether the last fix worked.

CFO Office — A real profit and loss from the till, the cost sheets and the books — reconciled. — CFO Office
A real profit and loss from the till, the cost sheets and the books — reconciled.

Every number opens — Click any figure and it shows the days behind it, then the locations, then the receipts. — Every number opens
Every figure carries its definition, its formula, the locations behind it and what moved it — down to the day and the receipt. This is the difference between a dashboard and something you can act on.

Photographed 10 September 2026 from the public demo — sample data, real software.

What you get

Eight ways it earns its keep every day.

Built for an operator who has ninety seconds, not ninety minutes.

☀️

The 90-second morning read

A per-store huddle: yesterday, this week vs plan, food and labor pace, and one focus for the day — live from Toast.

🎯

Command center + sentinel

Exceptions first, capped so it's never noise: budget misses, food creep, labor pacing into OT.

🍽️

Product-mix intelligence

Every item ranked by units, revenue and % of mix, with movers and an AI read on concentration and pricing.

📦

Where every order comes from

🤖

An operator that drafts

The AI drafts the renegotiation email, the schedule trim, the next move — behind your one-click approve gate.

✅

Reconciliation & report cards

Toast vs your books reconciled, A–F store report cards, and a forecast that's backtested — not guessed.

🌦️

A forecast that knows it rained — and that the parade is Sunday

Weather, holidays, paydays, school breaks, street closures, your own promotions and closures. Each store's own measured response to each one moves tomorrow's forecast and the labor plan behind it — and the screen says why.

🧾

Recipe & menu costing

What each plate costs to put out, from your item costs and your mix — so the menu is ranked by the margin it earns, not the price on the board. Per-item purchase prices join when your invoices are connected.

The part nobody else does

It tells you when it doesn’t know.

Every number on every screen carries its own coverage. A missed sync, a sheet nobody filled, a restaurant that was closed, a feed you have not connected yet — each is shown as an absence, with the reason, and never as a zero pretending to be a measurement.

It is the least glamorous thing in the product and the reason the rest of it is worth reading. A dashboard that prints $0 for money it never looked at has taught you not to trust the numbers beside it.

Net sales · yesterday$14,208

Food cost · this week28.4%

Deposited · yesterdaynot measured — no bank feed connected

Labor · Sundaynot measured — the sync missed that day

Prime cost · this week58.1%

What another dashboard would show as $0 and 0.0%.

Every number, examined

Every number says what it is and how sure it is.

Not a footnote and not a tooltip. It is written on the card, in words, before you decide anything on it.

A dashboard number is usually just a number. Here every KPI card carries three more things: what kind of number it is — a measured total, a rate, a share, an estimate; how sure it is — measured, estimated, or incomplete, with the reason; and why it moved — the part of the formula that did the moving, in one sentence.

Measured, estimated or incomplete. A figure built from a source that has not posted says so. Labor before payroll lands is an estimate from Toast hours and is labelled as one.
A week still trading is paced, and only on its closed days. Running totals project to a full week; rates do not, because a rate is already a rate. A finished week is compared, never projected.
Which part moved. Food cost rose because the numerator rose or the denominator fell, and the card says which, with the dollars. A mix change names the top movers and their share.
Nothing is guessed. A card that cannot say why says it needs more weeks, and how many.

Labor cost · this week

27.8%

estimateda rate4 of 7 days closed

Estimated from Toast hours at each store’s own recent cost per hour, until payroll posts. Up 1.4 points on last week: hours rose 6% while sales held flat, so the numerator did the moving. Not paced — a rate is already a rate.

An example card, worded the way the product words it — sample figures.

The report card

Five KPIs, graded separately. Then one letter.

Every store, every week. The letter is built the way a good operator judges a week — against the target you set, with history as context, and never on a week that has not finished.

Sales, food cost, labor cost, sales per labor hour and labor cost per hour are each graded on their own — against your target first, then last year, the last thirteen weeks, the last four and last week — and weighted into one letter for the store and one for the company. Every letter opens: the baseline each KPI missed, by how much, and the direction of travel kept as a separate arrow.

The letter is your target, not your past. A store that improved from a bad base is not an A for it, and one that never moved is not a C for standing still.
Meeting your target can never grade below a C. History adds context; it cannot pull a hit target under the line.
Nothing is graded on incomplete data. A week still in progress is not graded at all. A KPI whose payroll or cost sheet has not posted is left ungraded and its weight moves to the others — a half-landed week never manufactures an F.
Labor is graded from Toast hours until payroll posts, at the store’s own recent cost per hour, and marked estimated until the real figure lands.
Baselines are built only from complete, plausible weeks. A quarter of a payroll posted is not a baseline; sanity bands keep it out.
Targets, weights and bands are yours — per company, and per store where a store differs.

How one letter is built

Read from the grading engine, not typed here — the same weights every screen grades with.

Salesvs target, then history30%of the letter

Food costlower is better20%of the letter

Labor costlower is better20%of the letter

Sales per labor hour15%of the letter

Labor cost per hourlower is better15%of the letter

Inside each KPI, the baselines weigh

Your target or budget70%the standard

Last year, same week10%context

Last thirteen weeks10%context

Last four weeks5%context

Last week5%context

ABCDFA from 90, B from 100, C from 100, D from 100 out of 100 — on target scores 75, six per cent better scores 90.

The report card — Five KPIs graded on their own against your target and your history, then weighted into one letter — with the baseline each one missed. — The report card
Five KPIs graded on their own against your target and your history, then weighted into one letter — with the baseline each one missed. Photographed from the public demo — sample data, real software.

Two screens no dashboard has

Where the cash went. Where the discounts go.

Cash Integrity walks every day from gross sales to the money collected and the cash due, step by step, and a step that was never measured says so rather than printing a zero. Exceptions indexes each server’s discount and void rate against that location’s own peers — POS ids only, never names — so it tells a manager where to look, not whom to blame.

Cash Integrity — From gross sales to the money collected and the cash due, step by step — a step never measured says so instead of printing $0. — Cash Integrity
From gross sales to the money collected and the cash due, step by step — a step never measured says so instead of printing $0.

Exceptions — Discount and void rates by server, indexed against that location's own peers. POS ids, never names; a question for a manager, never an accusation. — Exceptions
Discount and void rates by server, indexed against that location's own peers. POS ids, never names; a question for a manager, never an accusation.

Photographed from the public demo — sample data, real software.

Ask your numbers

Ask from Claude, ChatGPT or any assistant you already use.

KPI360 is an MCP server. One token per company, made in Settings, and your assistant can list your locations, read the latest complete week’s KPIs against budget and last year, see the week’s weather outlook with its measured effect, and put a question to the CFO analyst in plain English. Read-only by construction: it explains and recommends, it never changes anything.

list_locationsget_kpisget_weather_outlookask_analyst

› which store’s food cost moved most last week, and why?

Union Square, up 2.1 points to 31.4%. Purchases rose $3,180 on flat sales; two protein invoices landed in the same week. The other five stores moved less than half a point.

ask_analyst · read-only · answered from the measured week, not a model’s memory

AI at the core

An AI CFO that reads your numbers —
and tells you what to do next.

And then it is held to it. Every action you approve is measured four weeks on against a sister-store baseline, and the outcome ledger states the result in dollars only where the arithmetic is clean. An action with no detectable effect is counted and said — never dollarized. A window still running is named, never estimated.

Every week, KPI360's AI reads your P&L, product mix, labor and channels, then writes the brief and drafts the exact moves. You approve; it acts. No dashboards to decode.

Weekly CFO brief

The five things that changed, why, and the dollar impact — in plain English.

An operator that drafts moves

Schedule cuts, 3P renegotiation emails, menu actions — queued behind your approval.

Ask anything

“Why was Riverside down last week?” — answered from your real numbers, instantly.

KPI360 AI · working sessionwriting

◆ Proposed action · needs approval

Cut 40–60 labor hours at Riverside next week and hold SPMH at $85+ before restoring shifts. Est. impact: +$2,300.

Your books, whichever ledger you keep them in

Your POS knows what you sold. Your ledger knows what you spent.

QuickBooks Onlinereads your reports · posts the nightly journal Restaurant365reads sales, labor and invoices · writes nothing today

Neither one can tell you whether you made money. KPI360 holds both sides at once — so the P&L, the cash position and the money that never landed all read from the same week.

Toast POS net sales

QuickBooks deposits

Mon$12,480$12,480matched
Tue$11,905$11,905matched
Wed$14,220$14,220matched
Thu$13,640$11,793−$1,847
Fri$19,310$19,310matched

Found in the gap

$1,847 of Thursday’s third-party sales never reached the bank. Commission was deducted twice on the delivery payout. Sitting in neither system alone — only in the join.

Operator P&Lweek to date

Revenue$71,555

Food28.1%

Labor31.7%

Occupancy9.4%

Other18.5%

Net12.3%

Rent, insurance and G&A come from your books. Sales, food and labor come from your POS. One statement, both sides.

The point of all of it

One place your CFO, CEO and COO actually open.

The CFO opens the close. The ops director opens the exceptions. The GM opens the morning read. Three jobs, one set of definitions — so nobody spends the meeting arguing about whose number is right.

POS, payroll, vendor invoices and the general ledger land in the same screens, mapped to the same locations, checked against each other — with an AI accountant that reads your books the way a CFO would: GAAP-grounded, citing your own figures, flagging what to do this week. Not another dashboard beside the stack. The place the stack reports to.

Your chart of accounts, readKPI360 learns your naming — not a template’s.

Books vs POS, tied outTwo systems counted the same week. The gap is stated, never smoothed.

AI findings, figures citedPrime discipline, cash vs profit, ageing receivables — every claim checkable.

Books AutopilotNew

Every night, KPI360 posts each restaurant’s daily sales journal entry into its own QuickBooks file — net sales, sales tax collected, tips, gratuity — in your bookkeeper’s own format, on the accounts your ledger already uses. No more hand-entering sales, no more months of catch-up. A month your bookkeeper has closed is never touched by a correction, every correction is previewed and reversible, and a post is create-only and duplicate-proof: it checks QuickBooks before every post, never edits, never deletes.

One switch. On when you’re ready, off any time — connecting your books never requires it.

Journal · KPI-0104posted 12:30am

Food Salescredit$9,842.00

Sales Tax Payablecredit$873.48

Tips Owedcredit$164.20

Undeposited Fundsdebit$10,879.68

Balanced · duplicate-checked✓

An operator’s P&L

Prime cost, occupancy and flow-through — not a 300-line chart of accounts. Same week, every location, side by side.

Cash in and out, from your books

What came in, what went out and what the books say is left — per company file, for the window you choose. The bank's own balance joins it when you connect the feed.

Who owes you, and who you owe

Receivables and payables by age, with the catering invoices that quietly slipped past sixty days pulled to the top.

Reads everything, writes only what you approve

Every report your books serve, on demand. It writes two kinds of thing and nothing else: the nightly sales journal, once you switch it on, and a correction you ask for — a missing account, a duplicate entry it posted itself, a reclass. Every one is previewed first, confirmed by typing, logged, and reversible.

The month-end close

Eleven facts per restaurant. Measured, not ticked.

Every close you have ever run was a checklist somebody ticked. This one is a measurement: each fact is either true of your data or it is not, and the ones that are waiting name the source they are waiting on.

Riverside · August9 of 11

Every day measured31 days read from the till
Sales journal posted31 days posted
Cash identity holds31 days checked, every one holds
No bill over 90 days2 bills, $4,180 — open A/P
Inventory countedtrue cost 29.4% of sales
Sales tax remittedAugust period, $8,412
Deposits matched to the bankwaiting on the bank feed

A tick is a fact the system measured. Waiting names the source that would decide it — never a box somebody forgot.

True cost of goods, from a real countOpening + purchases − closing, per restaurant per month. Not the food cost off an invoice total — what the kitchen actually used.

The cash ladder, rung by rungGross to discounts to comps to refunds to net, then tax, tips and charges to what was collected and what is owed as cash. Every rung measured or shown absent.

Sales tax by filing periodWhat the till collected for the state, grouped into your filing periods. A period that has ended is never called late on a figure nobody measured.

What the kitchen should have usedEvery plate sold × its recipe, against what the count says was used. Stated only when the recipes cover the menu — a half-covered comparison measures the recipe book, not the kitchen.

The payment run proposes

Pick the bills, and KPI360 shows the run oldest-first against your cash cover — the bank, after payroll, after the reserve. Record this run writes the decision down: these bills, this date, by whom. Then it opens each bill where the money actually moves.

KPI360 never moves money. It proposes, records and opens — you release the payment in QuickBooks Bill Pay or Bill.com. Nothing on this platform can send a payment, and nothing is scheduled on your behalf.

Due in the next 7 days$18,940

Cash cover after payroll$26,300

Shortfallnone

Record this run →

Opens each bill in your own system to release

What it connects to

Everything it reads. The little it writes.

One table, kept current. A connection that is built but waiting on a key from you says so rather than appearing as though it were running.

Connection	What it reads	What it writes	Today
Toast POS	Orders, items, labor hours, dining options, tenders — 24 months on connection, then every one to two minutes. The history checks itself: every week’s two records are compared and re-read when they disagree, and an option Toast can no longer name is named once and kept.	Nothing	Live
QuickBooks Online	Profit & loss, balance sheet, cash flow, A/P and A/R ageing, deposits, vendors, your chart of accounts — one company file per restaurant	The nightly sales journal, once you switch it on · corrections you approve, each previewed and reversible	Live
Restaurant365	Daily sales, labor punches, vendor invoices. Your R365 locations are matched to your restaurants before a single number is used — where the names leave doubt, KPI360 asks rather than guessing. Access is issued through your Customer Success Manager, and the credential is tested before it is stored	Nothing today	Live
Google Sheets	Cost sheets, payroll, budget and last year — however your sheets are laid out	Nothing	Live
Weather & your calendar	Each restaurant’s own weather, US holidays, paydays, school breaks, street closures permitted on its block	Nothing	Live
Bill.com	Vendors and open bills, so A/P shows every bill once whichever system it came from	Nothing — you release payments in Bill.com	Built · waiting on your keys
Plaid	Your bank’s own balances, cash deposits and debits — the half of the cash ladder nothing else can measure	Nothing	Built · waiting on your keys

KPI360 never moves money and never sends anything on your behalf. It reads six systems and writes two kinds of entry into one of them, both of which you switch on and can switch off.

Your concept, not ours

Paste your website. It becomes your app.

KPI360 reads your logo and colors and re-skins the entire platform to match — light and dark. Every concept you run gets its own branded command center on its own subdomain, with fully isolated data. One operator, or fifty.

Brand it in a minute →

SCopper Fork

One theme engine · every concept, its own look — pulled straight from the brand's site.

Built for groups & holding companies

One login. Every brand you run.

Operate ten concepts under one umbrella? Onboard each brand once — its own POS or spreadsheet, its own look, its own isolated data. Then read the whole portfolio consolidated, or drill into a single brand in one click.

Portfolio roll-up. Net sales, prime cost and labor across every brand, side by side and weighted.
Per-brand isolation. Each concept’s team sees only their own numbers — never another brand’s.
Admin that flexes. See all brands or one, from a single account — with brand-level admins you invite.

Onboard your group →

Atlas Restaurant Group4 brands · 18 locations · portfolio view

SCopper Fork6 locations$1.18M

NNixta Taqueria4 locations$980k

MMarea Coastal5 locations$1.02M

PPinault Bakehouse3 locations$760k

Portfolio net sales · this week$3.94M

The math

Restaurants lose 4–8% to undetected drift. You only need back half a point.

That 4–8% figure is the National Restaurant Association's — food-cost variance nobody caught, labor drifting a point, sales data nobody read. KPI360 exists to catch it. The calculator below assumes you recover just 0.5% — the conservative end of what operators report.

Number of locations Avg weekly sales per location

Sources: NRA 2025 Operations Report (4–8% lost to undetected variances). In a survey of ~10,000 US locations, 61% of operators using AI analytics reported reduced food costs and 62% reduced labor costs.

$62,400

recovered per year at just 0.5% of sales

KPI360 for the year

What it recovers

What restaurant analytics costs elsewhere · published & reported 2026 pricing, per location per month

$499–749

Restaurant365

$350

MarginEdge

$400+

Avero

$150–400

Tenzo

$99

KPI360 — every feature

Different products solve different problems — several of these are excellent at invoices and inventory. KPI360's job is the executive layer: what changed, why, what to do, and what doing it was worth — measured. At $99 it is priced so every location can afford to know.

Already on Toast?

Toast tells you what happened. KPI360 tells you what to do.

Keep Toast's built-in reports — they're good, and KPI360 reads from the same source. The difference is the layer on top: not another dashboard, an analyst.

Toast's reports — free, descriptive

Real-time sales, orders and labor dashboards
Custom report widgets and exports
One location at a time, one day at a time
You read the numbers and decide what matters

KPI360 on top — prescriptive

Flags the two or three things costing you money — with a dollar figure
Every location and every week side by side, vs last year and vs budget
A–F report cards, drift alerts, and peer benchmarks
The AI drafts the fix — you approve it in one click, and the result is measured

Read-only on top of your existing Toast — a five-minute connect, and nothing about service changes.

How we compare

Feature by feature, against the tools you have heard of.

Eleven things a multi-unit operator actually needs from the numbers, and who offers them. Full means it is there and works without a data team. Partial means it exists with real caveats: an add-on, something you have to build, or a narrower version.

FullPartialNot offered

Capability	KPI360	Tenzo	Restaurant365	Avero	Power BI	Toast
Restaurant-native metric definitions	Full	Full	Full	Full	Not offered	Full
Weekly prime cost vs benchmark band	Full	Partial	Full	Partial	Not offered	Partial
Cash due vs deposited reconciliationmeasured both sides: cash due per restaurant per day from the till, deposited from your bank feed — connect the bank and the ladder closes to the penny	Full	Not offered	Partial	Not offered	Not offered	Partial
Sales bridge: traffic, check, mix, price	Full	Partial	Partial	Partial	Full	Not offered
Menu engineering on margin dollars	Full	Partial	Full	Full	Not offered	Partial
Anomaly flag with named driver	Full	Partial	Not offered	Not offered	Full	Not offered
Void, comp and discount exceptions	Full	Partial	Partial	Full	Not offered	Partial
Receipt-level drill from any KPI	Full	Partial	Full	Full	Full	Full
Demand forecasting on weather, events and your own calendar	Full	Full	Full	Partial	Full	Partial
Usable without a data team	Full	Full	Not offered	Full	Not offered	Full
Published, predictable pricing	Full	Not offered	Partial	Not offered	Partial	Full

How the forecast works — the row above, in full

Most forecasting features are a rule someone wrote down: rain is bad for lunch, a holiday is good for dinner. KPI360 does not assume any of that. It measures each effect against your own sales, one restaurant at a time, and applies it only where it is real.

It reads what is coming. The weather at each store, US holidays, paydays, school breaks from any calendar you follow, street closures and parades permitted on that store's own block, and your own promotions, private events and closures. Concerts and games nearby join when you add a ticketing key.
It measures each one against your history. A parade day is compared with the same weekday in the surrounding weeks — so a summer street fair is judged against summer, never against February. Under six occurrences, or under five percent, it stays context and moves nothing.
It applies at most three, and shows its working. Every adjusted day names the measurement behind it, in words: Labor Day × measured −71% (12 holiday days in your history). A day you mark closed is zero, and the labor plan follows the forecast down.

Your own numbers, on your own calendar, with the arithmetic in the open. That is the difference between a forecast you can schedule against and one you argue with.

The other columns reflect each product's published capabilities as we read them on 3 September 2026, and we keep them current. KPI360's column is what ships today, not a roadmap. If we have one wrong, tell us and we will correct it.

Questions

The things operators ask first.

Do I need to change how my restaurants work? +

No. KPI360 reads from Toast and your existing sheets — read-only. It writes to your books in two cases only: the optional nightly sales journal, after you switch it on, and a correction you ask it to make — each previewed, confirmed by typing, written to the record of changes, and reversible. Your team keeps working exactly as they do; the platform gives you one live view on top.

I already get reports from Toast — why add KPI360? +

Keep them — KPI360 reads the same data. Toast's dashboards are descriptive: they show you the numbers. KPI360 is prescriptive: it compares every location against last year, budget and its own baseline, flags the few things that are actually costing you money with a dollar figure attached, and drafts the fix for your approval. It's the difference between a report and an analyst.

What if I'm not on Toast? +

Drop in any spreadsheet — a Google Sheet or CSV export from your POS, back-office or accountant. Our AI reads your headers and sample rows and maps them to KPI360's fields; you confirm, and your dashboards populate. No fixed template, no Toast required.

Can I run several brands under one account? +

Yes. A holding company can onboard each brand once — its own data source, branding and isolated workspace — and admins see the whole portfolio consolidated or drill into a single brand. Brand teams only ever see their own numbers.

How does billing work? +

$99 per connected POS location per month — every feature included, no tiers, no free-trial expiry games. We detect how many locations you connect and bill only for those. Add or remove a location, the bill adjusts.

How fast can we be live? +

Same day. Connect Toast and your sheets, paste your website to brand it, and your command center is populated within minutes — smarter every week as the AI learns your patterns.

Does the AI do anything on its own? +

Never without you. The operator agent drafts actions and nothing happens until you approve it with one click. The one exception is the one you explicitly switch on: Books Autopilot posts each day’s sales journal nightly — off by default, one switch, create-only and duplicate-proof, and reviewable line by line on the Journal Entries screen.

Is my data secure and separated? +

Every tenant's data lives in its own isolated namespace, limited to the team you invite. One concept never sees another's numbers.