Concise Detailed
The Kettle BlackAn approach note by The Kettle Black

Flaky Pastry Croissant Co. · Dubai · July 2026

Every gram, every dirham, accounted for.

You already weigh butter to the gram when you laminate. This brings the same precision to the business — reconciling what the kitchen makes with what the counter sells, every night, across every store, so the leaks show themselves before month-end does.

For
Flaky Pastry · 4 outlets, 2 opening
Built on
Your own Sapaad POS data
Spirit
Weigh everything · waste nothing

A note for the Flaky Pastry family — from one conversation, before any data.

Know tonight what tomorrow's chocolate should weigh
See leakage per store, per ingredient, per day
A COGS your accounts team can finally explain

Dubai's first croissanterie · precision is the house style

The Idea

Two ledgers, one truth.

Every bakery runs on two stories that never quite agree. The pull story: what the POS says you sold, exploded through your recipes into what should have been consumed. The push story: what the kitchen actually drew, batched, froze, filled and threw. Today those two stories only meet at a painful month-end stock take. We make them meet every night — store by store, ingredient by ingredient — and the difference between them gets a name, a weight and a dirham value.

It is not about policing the kitchen. It is a scoreboard, never a blame sheet — the numbers do the reminding, so your people can do the baking.

Nothing to rip out
Built on data you already own
Sapaad stays exactly where it is — the POS, the purchase orders, the requisitions. We take the dump (an Excel export every 6, 12 or 24 hours is enough) and build the intelligence layer on top. No integration project, no retraining the floor.
The engine
Recipes are the backbone
Raw ingredients → sub-recipes (your chocolate cream, praline, ganache, shakshuka sauce) → finished products → consumables riding along (a straw with every cold drink, a box with every delivery). One explosion engine turns every sale into grams — and every gram into dirhams.
Human first
A prompt, never a verdict
A 7% variance is a question, not an accusation — the recipe might be wrong, the scale might be off, the supplier might have shorted a kilo. The system points to where to look; your people decide. Hit 3% and the scorecard says so, loudly.
4 + 2
outlets today, two opening — small seepages become big numbers at this scale
Nightly
pull vs push reconciliation, instead of a stock take every 10 days that still doesn't match
~20 items
high-value ingredients — chocolate, butter, pistachio — counted daily in ten minutes
3–4%
your own wastage estimate. Even one point recovered pays for this many times over
One Life Cycle

From purchase order to plate — with every leak marked.

Inventory doesn't vanish in one place. It slips a little at each of eight steps. The system watches all eight — so when the number is off, you know which step to ask about, not just that "something's wrong."

1Order
POs to every supplier, in their pack sizes — butter by the 10-kg pack, chocolate by the 3-kg bar, basil by the pot.
Leak: ordering on habit, not on cover-days.
2Receive
Photograph the paper invoice — AI reads it and reconciles. Ordered 10 kg, invoice says 9 — the system books 9 and flags the gap.
Leak: "received" punched as ordered, not as delivered.
3Warehouse
Bulk stock — 50,000 sleeves, 5,000 water bottles — with cover-days and reorder points per item.
Leak: requisitions accepted unconditionally, no trend check.
4Transfer
Store requests → warehouse accepts → driver moves it → store confirms. Both ends punch a count; the two counts must agree.
Leak: sent 500, received 480, nobody noticed.
5Produce
The push. Batches of chocolate cream, ganache, praline — planned from forecast, weighed out, part used, part frozen with a date.
Leak: 520 g scooped where the recipe says 500.
6Sell
The pull. Every bill on Sapaad — dine-in, takeaway, delivery — explodes through recipes into theoretical consumption, consumables included.
Leak: assuming what's pulled is what was consumed.
7Close
The display rule — a bakery can't look empty at 10:30 pm. What's thrown at close is logged per SKU, per store, per night.
Leak: the 9:40 pm bake nobody was going to buy.
8Count
Ten minutes, ~20 high-value items, every night: the scale says 37.9 kg, the ledger expected 41.2. The gap surfaces tomorrow, not at month-end.
Leak: monthly counts that arrive too late to ask anyone anything.

Theoretical stock is a promise. The nightly count is the truth. The whole product is the space between the two.

How It Helps

Three leaks, caught in the act.

Illustrative moments — the shape of what surfaces once your two months of data are in.

The 3.3 kilos of dark chocolate
Dark 55% is your most expensive everyday ingredient — it comes in 3-kg bars and goes into half the display. After 9 days, the ledger expected 41.2 kg on hand; the scale said 37.9. That's 3.3 kg — roughly AED 460 — gone quietly, in one store, in nine days. Now it's a daily ten-minute ritual with an expected weight printed next to the scale, and the gap has to explain itself while the week is still warm.
The 70,000 wet wipes
A store's monthly requisition comes in at 70,000 wipes. Its six-month average is 50,500 — and its covers haven't grown. Today the warehouse accepts unconditionally; with the trend on screen, the warehouse taps "ask why" before a single box moves. Consumables become ingredients: wipes per cover, straws per cold drink, boxes per delivery order — each with its own expected number.
The 9:40 pm bake
The display must look generous at 10:30 — that's the brand, and it's non-negotiable. But the close-out log shows that on the last six Tuesdays, 17 of the 24 almond croissants baked after 9 pm were thrown. The classic still sells at 10:15; the almond doesn't. Same full display, one tray less of the right SKU — the look stays, the waste doesn't.
Ask in Plain English

No reports to commission. Just ask.

The proprietor, the warehouse, the accountant — each asks in their own words, and the answer comes back with the evidence behind it.

Which store's variance is worst this month — and which three ingredients are driving it?
What should the dark chocolate at City Walk weigh tonight?
Why is food cost at 27.4% when my paper costing says 25?
What did we throw at close last week, store by store — and which SKUs should bake one tray less?
Given the cold drinks we sold in May, how many straws should have left the warehouse?
Which products earn the most margin per tray — and which are just taking up display space?
In Practice

Five roles, five windows — one set of numbers.

Everyone sees the same truth, cut for their job. Access is scoped: the kitchen sees the kitchen, the warehouse sees the warehouse — and the proprietor's god mode sees everything, everywhere.

The Proprietor God mode
Every store, every ingredient, one leaderboard. Opens the phone at 7 am to yesterday's variance per store, the dirham value of the week's leaks, and one suggested question to ask each manager today. Sets the targets and the incentive lines: hit 3%, it shows in green.
The Kitchen Kitchen access
Today's production plan from the forecast, batch cards for every sub-recipe, and the nightly ten-minute count — twenty high-value items with the expected weight already printed. Punch what the scale says; done.
The Warehouse Warehouse access
One screen: requisitions in — approved with six months of trend on screen — transfers out, both ends counted. Cover-days per item and reorder points that account for supplier lead time, so the boxes never quietly run out, and never quietly pile up.
The Store Own store
Covers, consumable ratios, the display close-out list, and their own store's variance — the score they're playing to. What gets seen daily gets managed daily.
The Accountant Finance access
A COGS that arrives explained — food, beverage, consumables, packaging, each with its bridge from plan to actual. Month-end becomes a review, not an investigation. Costs and selling prices editable in one place.
A Peek Inside

One store, one night — through four lenses.

City Walk, after close on a Thursday. The same numbers, cut for four different jobs. Every note is a specific signal — a weight, a count, a dirham value — drawn from a named record, with a clear "what you can do." Try the tabs. (Illustrative figures — your two months of data make these real.)

Every Gram · City Walk · Thursday close
CW
City Walk
Dine-in · takeaway · delivery  ·  reconciled nightly against production & counts
Variance 1.8% ↓Food cost 27.4%412 coversClose-out AED 318
What the proprietor sees across all four stores, before the first coffee.
The leaderboard — variance by store, this month
City Walk
1.8%
Boulevard
2.6%
Marina
3.1%
DIFC
6.8%
DIFC's gap is 80% dark chocolate and butter sheets — two ingredients, not twenty.
What you can do
One question for DIFC today: "walk me through how the chocolate is weighed out" — not an audit, a conversation.
Nightly pull–push reconciliation · all stores
The expensive one
Dark 55% at City Walk: ledger expected 41.2 kg, last night's count says 37.9 kg. Nine days, 3.3 kg ≈ AED 460.
What you can do
It's on the daily count list from tonight — the gap can never grow older than a day again.
High-value count register · recipe explosion
Butter, the backbone
Croissants sold + close-out waste needed 62 packs of sheet butter this month. The kitchen drew 66 — a 6.5% over-draw, worth AED 1,140/month at this pace.
What you can do
Spot-weigh one lamination batch this week — if the recipe's wrong, fix the recipe; if the scale's wrong, fix the scale.
Production draws vs POS sales · butter sheets
"Twenty grams over doesn't change the flavour."— every kitchen, everywhere. True for the croissant; expensive for the company.
What you can do
Put the scoreboard where the kitchen can see it, and pay the 3% club a bonus. Behaviour follows the number it can see.
Proof it moves
Since the dashboard went up, DIFC has come from 6.8% toward 4.1% in three weeks — no memo, no meeting. Just a visible number and an incentive.
Variance trend · 21 days
Display economics
Close-out waste last night: AED 318 across the display — 61% of it from three SKUs baked after 9 pm.
What you can do
Approve the suggested late-bake plan: one tray less of almond and pistachio after 9, classic unchanged. Display stays full.
Close-out log · 6-week Tuesday pattern
Tonight's count — ten minutes, twenty items
Dark chocolate 55% (3-kg bars)expect 37.9 kg
Sheet butter (10-kg packs)expect 4.1 packs
Pistachio pasteexpect 6.2 kg
Vanilla podsexpect 0.42 kg
+ 16 more on the register
What you can do
Put what the scale says next to what the ledger expects. Match within tolerance — done, green, good night.
High-value register · expected from recipe explosion
Batch card #214 · chocolate cream
Planned 5.0 kg (50 fills @ 100 g). Sold 47, threw 3 — the batch should be gone. The bowl scaled out at −400 g early: filling is running ~8 g heavy per croissant.
What you can do
Check the piping nozzle dose against the 100 g spec — one adjustment saves a batch a week.
Production log · batch vs POS fills
Today's production plan
Bake 140 classic · 65 almond · 40 pistachio · 30 za'atar — from four weeks of this weekday's sales, plus the display buffer. Enough for a full case at 10:30, without a bin full at 11.
What you can do
Adjust and confirm — your number wins, the forecast just does the arithmetic first.
Sales forecast · trailing 4 weeks by weekday
In the freezer
Chocolate cream: 12.4 kg across 3 batches — oldest is 11 days. Praline: 4.8 kg, 6 days. Use-first flags are on the oldest tubs.
What you can do
Pull batch #202 first today — and skip making a new batch until Thursday.
Sub-recipe batch ledger · FIFO ageing
"Recipes are the backbone — the costing and the selling price both stand on them."— from our first conversation
What you can do
A recipe editor with your real grammages — ideal cost and operational cost side by side, so "the recipe is wrong" stops being a guess.
Your score this week
Kitchen variance: 1.8% — inside the 3% line, four weeks running. That's the bonus threshold.
Store scorecard · incentive tracker
Requisition #R-1188 · needs a look
City Walk asks for 70,000 wet wipes. Their six-month average is 50,500 — and covers are flat. That's 39% above trend, worth about AED 1,900.
What you can do
One tap: "ask why" — the request pauses with a note, instead of moving unconditionally like today.
Requisition history · store trend, 6 months
Transfer mismatch
Sent 500 sleeves to City Walk; store confirmed 480. Twenty sleeves are in a van, a miscount, or nowhere.
What you can do
Both ends punch a count now — mismatches surface the same day, while the driver still remembers the run.
Transfer log · dispatch vs receipt
Reorder now, not soon
Croissant boxes: 11 days of cover left at current pace — supplier lead time is 15 days. The gap is four days of empty shelves.
What you can do
The PO for 50,000 is already drafted — review and send. Reorder points now include lead time, not vibes.
Cover-days · consumption pace vs lead time
Straws, reconciled
Issued 4,120 straws this month; the POS sold 3,610 cold drinks. Every cold drink now carries a straw in its recipe — so 510 straws are a question, not a rounding error.
What you can do
Consumables are ingredients now — wipes per cover, boxes per delivery order — each with an expected count.
Warehouse issues vs POS channel sales
Still water, at a glance
Warehouse holds 3,900 bottles; stores hold ~610; selling pace 260/day → 17 days of cover, comfortable.
Stock by location · sales pace
Receiving, reconciled by photo
Yesterday's butter delivery: PO said 10 packs, the invoice photo reads 9. Booked 9; credit note flagged automatically.
What you can do
Snap the paper invoice at the door — the system does the comparing, not the busiest person in the building.
PO ↔ invoice scan ↔ goods receipt
The COGS bridge — plan 25.0% → actual 27.4%
Chocolate & butter variance+0.9 pt
Display close-out waste+0.6 pt
Receiving gaps (short deliveries booked full)+0.5 pt
Consumables over-issue+0.4 pt
2.4 points of drift — explained line by line, for the first time.
What you can do
Each line has an owner and a fix already on this page — the P&L conversation becomes four small actions, not one big mystery.
Pull–push ledger + close-out log + receiving reconciliation
Consumables check
Consumables at 1.31% of revenue vs the 1.0% budget line. Wet wipes alone are 0.19 of the 0.31 excess — one store, one habit.
What you can do
The requisition flag on the Warehouse tab is the fix — this line just proves it in dirhams.
Consumables issues vs revenue · by store
Month-end, on the 1st
Stock valuation drafted by 9 am: AED 184,300 across four stores + warehouse — priced at latest invoice cost, with the ten largest movements annotated.
What you can do
Month-end becomes a review, not an investigation — the explaining happened nightly, all month.
Perpetual inventory · latest-cost valuation
Supplier price creep
Sheet butter is up 9% over three invoices; the croissant's ideal cost moved from AED 3.85 to 4.12 and nobody re-priced anything.
What you can do
Margin alerts per product — when an ingredient moves, every affected selling price gets a nudge, not a surprise at quarter-end.
Invoice cost history · recipe re-costing
"Our accounts team spends a tremendous amount of time and still can't properly explain the COGS booking."— from our first conversation
What you can do
Keep Sapaad's costs as-is — you get an interface to correct costs yourself, and the bridge recalculates instantly.
Channel margins, honestly
Delivery orders carry AED 2.10 of packaging each — dine-in carries AED 0.35 of consumables. Delivery margin is 4.1 points thinner than the menu suggests.
What you can do
See margin by channel, after packaging — then decide delivery pricing with real numbers.
POS channel split · consumable linkage
The Full Kitchen

Everything the data can cook. TBC — subject to data assessment

The dashboard above is one night. Underneath it sits a full analytical kitchen — each dish answering one question you're asking today, all from the same Sapaad dump. The final list is confirmed once we've assessed what your two months of data actually carries.

AnalysisThe question it answersBuilt from
Leakage & varianceWhere is the gap between what we should have used and what we did — per store, per ingredient, per day, in dirhams?POS sales × recipes vs production draws & counts
Menu engineeringWhich products are stars (high margin, high volume), which are display decoration, and which quietly lose money per tray?Item sales × ideal cost × close-out waste
Ideal vs operational costWhat should this croissant cost on paper — and what does it actually cost the way we really make it?Recipe grammages vs actual batch draws
Display close-out optimizerHow full can the 10:30 pm display stay while throwing the least — SKU by SKU, weekday by weekday?Hourly sales curve × close-out log
Consumable ratiosWipes per cover, straws per cold drink, boxes per delivery — is any store drifting from its own normal?Warehouse issues × POS covers & channel mix
Requisition anomaliesWhich store is asking for 39% more than its trend — before the warehouse says yes?Requisition history × sales pace
Receiving reconciliationDid we book what was ordered, or what actually arrived? Photo the invoice; the system compares all three.PO ↔ invoice scan ↔ goods receipt
Daily high-value countsAre the twenty expensive items — chocolate, butter, pistachio — where the ledger says they should be, every single night?Count register × expected-on-hand
Batch yield & freezer ageingDid the 5-kg batch fill 50 croissants or 46 — and which frozen tub must move first?Sub-recipe batch ledger · FIFO flags
COGS bridgeFood, beverage, consumables, packaging — plan vs actual, with every point of drift attributed to a cause and an owner.All ledgers, rolled up nightly
Price-creep & margin alertsWhich supplier price moved, which recipes it touched, and which selling prices now need a look?Invoice cost history × recipe tree
Production forecastingHow many of each SKU should tomorrow's bake be — full display at close, minimum bin at 11?Trailing sales by weekday · seasonal launches
Store scorecards & incentivesWho's inside the 3% line this month — and is the bonus doing its job?Variance trend × targets you set
One source of truth: your Sapaad dump Costs & selling prices editable by you New stores onboard by copying a store profile Seasonal launches slot into recipes & forecast
The Heart of It

Precision is already the house style.

Flaky Pastry was built on doing one thing with unreasonable care — lamination in full view, the finest ingredients, a display that never looks tired. The kitchen already lives by the gram. This simply extends that same craft to the part of the business nobody can see through a glass counter: where the grams go after they're bought, and before they're sold.

ForTodayWith Every Gram
The proprietorFeels the leaks; can't name themSees each leak with a weight, a store and a dirham value — daily
The kitchenMonth-end counts that never matchA ten-minute nightly ritual, and a scorecard worth a bonus
The warehouseAccepts every requisition unconditionallyApproves with six months of trend on screen
The accountantWeeks explaining a COGS that won't reconcileA bridge that arrives explained on the 1st
The stock takeEvery 10 days, all items, still wrongDaily for the expensive twenty; monthly for the rest
The next two storesTwo more places for the same seepageOpen on the system from day one — measured from the first croissant

You can't manage what you only count once a month. Count the twenty things that matter every night, and the rest manages itself.

The First Step

Two months of data. One working mockup.

Exactly as we said on the call: no commitment, no integration project. Send the Sapaad dump — or temporary access, and we extract it ourselves — and we return a working model on your real numbers. If it adds value, then we talk commercials.

From the POS
Sales & channels
Item-level sales with quantities — dine-in, takeaway, delivery — plus bills and covers. Two months, all four outlets.
The backbone
Recipes & ingredients
The ingredient master (pack sizes, suppliers, costs), every recipe and sub-recipe with grammages, and current selling prices. Costs imperfect? Fine — you'll get an interface to correct them.
The movements
Stock & flows
Purchase orders & receipts, warehouse requisitions & transfers, production entries, wastage entries, and the last few stock counts — warehouse and stores.
Excel export every 6 / 12 / 24 hours is enough Sapaad stays — nothing replaced Mockup first · commercials only if it earns them
Flaky Pastry Croissant Co. A Gentle Close

Weigh everything. Waste nothing.

Dubai's first croissanterie earned its name by showing its craft through glass — nothing hidden, everything done properly. Every Gram is the same idea, turned inward: the whole life of every ingredient, visible from the purchase order to the last croissant on the 10:30 display. The craft is already there. Now the numbers get to match it.

An approach note for Flaky Pastry Croissant Co. · July 2026

Flaky Pastry · Every Gram
The Kettle BlackAn AI product by The Kettle Black.