Five jobs Dyner took off the Sunday list

Every independent restaurant has a Sunday. Not the service, the other one: the four hours after the last table leaves, at the desk in the corner with the invoice folder, a stock sheet somebody has already spilled something on, and the till report. It is the only stretch of the week when nothing is being asked of you, so it is when the numbers get done.

That afternoon is what Dyner is aimed at. Not at making it prettier - at deleting the parts of it that are transcription rather than judgement. Five pieces, and what each one takes off the desk.

1. Reading the paperwork

Invoices, credit notes, supplier statements and stock sheets come in, in whatever format the supplier chose. They are read, matched and costed without anyone retyping a line.

Matching is the hard part and the part that gets glossed over in demos. Your supplier writes CHKN THGH BNLS SKNLS 5KG. You call it deboned chicken thigh and you count it in kilograms. Getting from the first to the second, every time, for every supplier, including the one who changed his product codes in April without telling anybody, is the difference between a system that saves you the Sunday and a system that gives you a new job of correcting it.

Once a line is matched it does not stop there. The ingredient's price updates, and every recipe and menu item built on it recosts, including the ones two levels down that only meet it through a sub-recipe. A supplier putting up one price should show as a thinner margin on everything made with that ingredient, the same day, without anybody remembering to go and look.

2. Stock that reconciles

A stock system that only records counts is a spreadsheet with a login. What makes one worth having is whether the period balances.

Four things move stock and most systems handle the easy two. Dyner handles all four: what you counted, what you threw away, what you sent to the other site, and what the kitchen turned into something else. That last one is the batch of demi-glace, the pre-portioned mince, the house pickles, and it is the one that wrecks a valuation when a system has no word for it.

Then the period has to close, and the closing valuation has to be derivable from where you started rather than merely printed beside it. Where the two do not meet, the difference is broken out ingredient by ingredient. Which means the conversation is about eight specific lines, and not about whether the software can be trusted.

A valuation you cannot defend line by line is a number, not a control.

3. Finding the money that leaves without a receipt

Theft in a restaurant is almost never a dramatic event. It is a void here, a staff discount that was never authorised, a refund at ten to midnight on a till nobody was watching, a no-sale opening the drawer between two cash transactions, the same item ringing up as a comp four Saturdays running.

Each of those is defensible on its own. The pattern is not, and a pattern is only visible if somebody is looking at every sale, every employee and every shift rather than at a monthly summary. Dyner does the looking and puts what it finds in a queue with an owner, a status and the underlying bill attached, so a flag can be opened, understood and closed.

You can add your own rules too: choose the metric, set the condition that should raise a flag, name it, switch it on. What a particular restaurant is actually worried about is rarely what a vendor guessed in advance.

4. A wage bill that gets checked before it is paid

Labour is the other half of prime cost and it is usually managed with a roster in one system, a clock-in book in another and payroll in a third, none of which are compared until the money has gone out.

Dyner puts rostered hours, clocked hours and what payroll is about to pay side by side while there is still time. A week rostered at 412 hours that ran at 438 is a conversation on the Wednesday and a cost on the Friday. Tips, attendance and employee performance sit in the same place, and payroll pushes through to SimplePay rather than being retyped into it.

5. Answers, without waiting for a report

Reports answer the questions somebody anticipated. The question you have at nine on a Tuesday is usually not one of them.

Athena takes it in ordinary English. Ask which dishes lost margin last month and you get the list, the chart, and the query that produced both, which is the part that matters: an answer you cannot check is an answer you have to take on faith, and nobody should be running a business on faith in a language model. Hand it a workbook or a PDF and it will read that too. Ask for something substantial and it will propose a plan before it starts.

The connections, which are the boring and important bit

None of the above works if the data has to be entered twice. Sales come from the till: GAAP, Lightspeed, Micros, Pilot Cloud, Munch, Restaurant Manager, Sally, POS Control, Universal, and Yoco's own kiosk and table products. Delivery sales come from the platforms. Purchases can come across from MarketMan. The finished cost of sales goes out to Xero or Sage against your own GL codes, and payroll to SimplePay.

Multiple sites are handled as a group rather than as several accounts you sign in and out of, which matters the moment you open the second one and start moving stock between them.

What we did not build

Dyner is not a point of sale and is not trying to become one. You already have a till, your staff already know it, and the last thing a kitchen needs is a new screen on the pass during service.

It is also not a general business intelligence tool with a restaurant theme. The difference shows in the vocabulary: a BI tool has dimensions and measures, and this has GRVs, wastage, sub-recipes and cash-ups, because those are the things that actually go wrong.

The Sunday afternoon does not disappear. But it stops being four hours of typing and becomes twenty minutes of deciding, which is the only part of it that was ever worth your time.