How it works

How we get our numbers, in 4 simple steps.

01 Cleaning your data

We start with your exports, as they are.

POS history, delivery statements and cost sheets arrive in different formats, with gaps and duplicates. We turn them into one clean record per item, location and week. Messy is fine.

Your exports3 systems, as received

POS2026-03-04,BURGER SIG ,8.95,,ST-184,dinename matched

POS04/03/26 | Signature Burger | 8,95 | store 184 | counterdate fixed

POS04/03/26 | Signature Burger | 8,95 | store 184 | counterduplicate removed

DeliveryPLATFORM;2026-03-04;Sig. Burger;10.45;184;fee 29%fee separated

POS2026-03-04,,8.95,,ST-184,name recovered

InvoiceINV 03/26 · Beef patty 150g · 1.12/unit · Store 184cost added

POS2026-03-04,FRIES LG,4.10,,ST-184,dinename matched

DeliveryPLATFORM;2026-03-04;Fries large;4.90;184;fee 29%fee separated

One clean recordDone

Item	Location	Channel	Price	Cost	Week
Signature burger	184	Dine in	8.95	3.41	W10
Signature burger	184	Delivery	10.45	3.41 + 29% fee	W10
Fries, large	184	Dine in	4.10	0.82	W10
Fries, large	184	Delivery	4.90	0.82 + 29% fee	W10

Duplicates removed. Names matched across systems. Costs added from the invoice. Fees split out by channel. We only ever need read access.

What goes in

Sales historyPast price changesSales by storeMenu mixWhat sells togetherDay of weekSeasonalityWeatherCommodity pricesIngredient costsDelivery feesYour rules and goals

02 The models

Six models, one question: what is this item worth, here?

Each one answers a single question and passes the answer on. Click a model to see what it does.

We always try the simple model first. A more complex one is only used if it predicts better on data it has never seen. See the sample letter →

The methods are standard statistics, nothing exotic. Ask us on the audit call and we'll walk you through them, line by line.

03 Guardrails

Your pricing rules, enforced.

Every price change the models propose has to clear sixteen rules we set with you in the first week. Then one of three things happens.

The models propose · exampleA price changeDry-rub wings +$0.80
Combo A +$1.20

Sixteen rulesChecked one by oneStep size, spacing, margin floor, price endings, the items your regulars watch…agreed with you in week one

Passes all sixteen · 78%+ confidentA recommendation in your letterDry-rub wings +$0.80: passed, goes in with its forecast.

Passes all sixteen · under 78%A test at a few stores firstNot sure enough to recommend, so it is measured before you roll it out.

Breaks one ruleStopped. It never reaches you.Combo A +$1.20: a 12.6% jump, more than the step-size rule allows.

Eight of the sixteen, in plain words

Step sizeNo single change bigger than a set amount or percentage.

SpacingA minimum gap between changes on the same item.

Menu shareOnly part of the menu changes in one cycle.

Inflation budgetA cap on how much the whole menu moves.

Price endingsPrices end the way your customers expect.

Threshold reviewA price crossing $10, $15 or $20 goes to you first.

Margin floorNo item drops below its minimum margin.

Anchor itemsThe items your regulars price-check are never raised.

Brand and franchise rules are added in the first week and enforced the same way →

Recommendations go to kitchens, not slide decks. They have to work during service.

04 The flywheel

It gets better with every result.

Every measured result goes back into the models. Your second letter is better than your first because your own stores taught it.

Every test is recorded. The change, the store, the result. Wins, losses and nulls alike.
Your own stores teach it. Everything the models learn about your chain comes from your measured results. Your data is never pooled with anyone else's unless you agree to it.

Talk with us

Want to see what your data says?

Send us one POS export and we'll show you where the margin is. The first look is free, and you keep the readout either way.

Talk with us rodrigo@pretaris.comalexander@pretaris.com