That is the thing a Kochi bakery has and no restaurant does: a case of finished goods sitting in front of the customer with a clock on it, and a diary of cakes promised for a day that has not arrived. This page is about the gap between what went in the oven, what got rung up and what went in the bin — and about the order book running alongside it.
The night shift proves and bakes, and by half past six the case is full of puffs, cutlets, cream buns, dilkush, rusk and the day’s bread — none of which anybody has asked for yet. What software can honestly do at that hour is narrower than this market pretends, so it goes first.
There is no production module here. Nothing asks the night shift to log a batch of two hundred puffs, nothing holds a batch record, and nothing reconciles a yield against what actually came out of the oven. Ingredients leave the shelf when a product is billed, not when it is baked, and the number in the case is a figure a person sets in the morning. Every competitor in bakery software opens with production planning. We would rather say this on the page than let you find it in week one.
What is real before opening is the recipe underneath. One sponge, one icing, one custard is authored once as its own item, and every gateau that draws on it points at that one — up to three levels deep. When the platform reads a product’s ingredient demand it expands the whole tree down to the raw goods, so a tray of pastries asks for flour, butter and eggs rather than for sponge. Each ingredient line carries its own wastage percent, and the effective quantity — the one costing and capacity read — is the quantity plus that wastage. The trim you lose squaring a sponge is priced in on the recipe instead of turning up as a hole at the end of the month.
The method itself sits with the product: steps grouped under their own headings, a yield label, a target time, a hold instruction, allergen and equipment tags, a target temperature and a flag on the step that must not be got wrong. It stays a draft until you publish it, org-wide or overridden at one branch. That sheet comes with Menu & Catalog, which is a core module, so it does not wait on Inventory & Recipes being switched on.
Nobody orders a puff by name. They point at the case, and whoever is on the till has to find that line before the queue behind them moves.
So the counter types a short code, not a name. One code stands for an item, a category, a section, a choice group, a choice option or an ingredient, in a single namespace across the whole business. The platform allocates one, or you set the one your counter already calls out. A retired code is never handed out twice, so a price list printed two years ago still resolves to the thing it named. And the code is copied onto the order line beside the name, so a receipt reprinted in March shows what was printed on the day.
The cake is the other half of that counter, and it is not a line you point at — it is a set of choices. Weight is a required single-select with a price delta on each option: 500 g, one kilo, one and a half. Egg or eggless is a second required group. Flavour is a third. Each group is built once and hung on every cake, and the required flag, the minimum and each option’s price can be overridden on one product alone — or the group switched off, or an option repriced, at one branch alone.
Set a count on a line at one branch and every channel spends the same number. Each order decrements it only if enough is left, an oversell is refused naming exactly how many remain, and a cancelled or rejected order puts the quantity back. Adding items to an order already placed is gated against that same count, because two more puffs handed over at the counter is a new sale. Where you switch it on, a product whose required ingredient runs out comes off sale at that branch by itself — including one that only touches the ingredient through a sub-recipe — and it never overwrites a switch you set by hand.
Above the case, the price board is a wall of numbers that changes as the shelves empty. A sold-out line is badged or dropped, prices are read off the live menu, and the screen is nudged to refetch when availability changes — including when a line came off on its own because an ingredient ran out. Veg marks are set per board, and the mark is three-state: veg, non-veg, or unmarked, where an unmarked item renders no mark rather than guessing, because that mark is a regulatory claim. The FSSAI licence prints in the footer from the branch, since it is issued per premises.
Saturday’s half-kilo butterscotch. The eggless one-kilo black forest for the Kakkanad office, name piped on top, collected at five. From November the diary fills with plum cake and becomes the month’s margin — and it is still one sheet of paper that a single illegible line can ruin.
An order can be taken for a future time to the minute. Your own counter staff can set any time from five minutes out, with no upper limit; a customer booking themselves must land between one hour and three days ahead. Opening hours are then checked against the collection time rather than the moment of booking — it is the one place a special-hours closure is enforced — so a booking for a day you have marked closed is refused there and then, with a message telling the customer to pick another slot. Scheduling is a collection mechanism and the platform treats it as one: it is refused on a dine-in bill and on an in-store QR order.
Thirty minutes before it is due, the branch gets one push and one operational event logged as needing attention. The mark is stamped even if the push fails, so it never nags twice, and an unpaid advance order never nags at all. The countdown before an unpaid order auto-cancels runs from the collection time, not from when it was booked, so a cake taken on Wednesday for Saturday is not cancelled on Wednesday evening.
The tea-time rush empties two shelves and leaves three untouched. Then somebody decides what to mark down and what to bin, and in most bakeries neither decision leaves a trace anywhere.
The markdown can stop being a decision and become a rule that is already running. A daily deal belongs to one branch, targets one line or a whole category — so ‘Bakes’ in a single row — runs on the weekdays you tick between a start time and an end time, and can be a percentage off, an amount off, buy X get Y, spend X for a free item, or buy X for a free item. There is an add-on shape too: while a bake is in the basket, a tea or a coffee is offered at a fixed price or a percentage off, capped at the qualifying quantity. Points on the bill matter more here than in most places, because a bakery’s customer is the same person four times a week.
The bin is a stock movement somebody types. Waste and expiry are recorded per item per branch with a reason that is required and capped at 300 characters, they always reduce stock whichever way the number is sent, and they are valued at the weighted average you actually paid — so the tray you threw out is priced at what it cost, not at what you hoped to sell it for. The waste report then lists every waste and expiry movement newest first, with the item, the quantity, the reason and the person who recorded it. Nothing sweeps the case at close on its own, and there is no shelf-life field on a cream bun. Somebody records it, and that is exactly why the figure is worth anything at the end of the month.
And then the cash, because a bakery counter is a cash counter. The float carries forward from last night’s count, the denomination breakdown has to total the counted cash before it will save, and the variance stays blank rather than reading zero until somebody has actually counted it.
Butter, cream, eggs, yeast and premix arrive weekly on trade credit and are priced by the kilo. A twelve per cent move in butter eats a season, and nothing in the shop says so out loud.
Start with the chiller. Expiry tracking is switched on per ingredient, the purchase line carries the batch number and expiry date you entered when it came in, and a check runs every six hours and pages the branch on anything expiring within seven days that still has stock on hand — and on anything at or below its reorder level. The inventory dashboard names the five items closest to expiry, each with its batch number and date. The cream with four days left on it reaches a screen before it reaches a nose. Be precise about what that is, though: it is the batch and date you entered on the way in, not lot tracking. Stock on hand is one quantity per item per branch, so there is no per-lot balance and no first-expiry order of consumption.
Then the count, which is scoped to a storage area — and the area types are the rooms you already walk. It can be blind, so whoever walks the chiller is not reading the book figure off the sheet in their hand, and it carries the moment the shelf was actually walked, so a crate of cream received while the count was being typed up does not read as a phantom shortage. Draft, submitted, posted; posting writes the correction as its own movement rather than editing a number quietly.
That consumption report is the reconciliation the month end never had: the difference between the flour that went out and the cakes that were sold stops being a feeling. It also gets a name. Seven detectors write into one queue with plain labels — possible stockout, unusual waste, price increase, waste above recipe, unexplained stock loss, stock value mismatch, low margin. Waste above recipe is the usage-variance one: what the recipes say should have left the shelf, against what actually left it. Unexplained stock loss is what a count found that the recorded waste does not account for. Each is acknowledged, explained in your own words, resolved or marked a false positive. Every threshold is yours: waste over ₹200, a supplier price up more than 10%, usage more than 15% off recipe, a count variance over 5% or ₹200, a margin under 20%.
Last, the Friday question — how much flour, how much butter. The forecast walks day by day to the date each ingredient runs out: a recency-weighted usage rate with last week counted triple and the week before double, a per-weekday factor that is only trusted after three observations and clamped between a quarter and four times, and a suggested quantity rounded up into whole purchase units, because you cannot order 2.3 crates. Items are bucketed as running out within one, two or three days, against a cover target you set — seven days unless you change it. There is no model in it, and nothing in it is AI. It needs Inventory & Recipes switched on and the forecasting flag on with it, and so does the exceptions queue above.
Nine mechanisms a bakery product is expected to have. These are the ones we do not have, with the reason rather than a shrug, because you will test every one of them in the first fortnight.
If one of those is the thing you actually need, say so — it is how we decide what to build next.
Switch on the ones this business needs and leave the rest. They write to the same record, so the recipe behind a gateau, the count on the chiller and the cake booked for Saturday are one set of numbers read from different screens.
One sponge standing behind nine gateaux, the trim priced into every ingredient line, and the bin at close recorded as a movement with a reason on it.
A recipe behind each dish, stock off the shelf as it is billed, margin on every plate.
The advance order book: a collection time to the minute, the opening-hours check run against that time, and the name to be piped printed under the line.
Punch the bill at the counter and the kitchen copy prints with it, on one order record.
The short code the till punches instead of reading a name off two hundred lines, the weight and eggless groups on every cake, and the method sheet behind the bake.
Categories, sections, choice groups and imports — one catalog every channel reads.
A count on the chiller, the freezer, the dry store, the prep area or the front counter — blind if you want it, and stamped with when the shelf was actually walked.
Purchase orders, supplier receiving and stock counts that reconcile to the shelf.
Waste above recipe and unexplained stock loss, raised into a queue a manager works rather than a report nobody opens.
Executive, location, menu, workforce and customer dashboards over one dataset.
The date the butter runs out, walked forward day by day, and an order quantity rounded up into whole purchase units.
Next week’s demand, and a flag on the number that does not look like the others.
The evening markdown written as a rule with weekdays and hours on it, and points for the customer who is in four times a week rather than four times a year.
Points on the bill, milestones, coupons, daily deals, and campaigns that announce them.
The order book on your own domain: the customer picks the collection slot, pays by UPI, and collects it at the counter.
Your own ordering site, on your domain: menu, online checkout and Express Pickup QR.
The price wall above the case, with sold-out lines dropped or badged and the branch’s FSSAI number in the footer.
TV boards built from your menu, and a screen list that says which TVs are actually on.
The other end of the same evening — the float, the denomination count, and a variance that stays blank until somebody has actually counted.
Count the drawer, log the expenses, then submit, approve and lock the day.
A case count held per branch, and a choice group switched off or an option repriced at one outlet on its own.
More than one branch, branch-level menu control, and routines published where you pick.
Tell us about your restaurant and the workflows you want to launch first.
We will take one dinner rush the whole way — ring a bill, print the KOT, call the token, mark the payment, dispatch the order, pay the agent and read the day close. On your menu, with your zones, in about thirty minutes.