MarginLab is a pricing engine for independent cafés. It shows what every GoFood and GrabFood order really leaves you, prices your menu with a demand model calibrated to your own sales, and rounds to prices that feel right — distilled into one decision you can act on this week.
Most cafés set prices by feel, copy the shop next door, and never revisit it. The margin you're missing is already on the menu.
Owners price from gut and competitor-glancing. Intuition is good at flavour, blind to elasticity.
Costs drift up; prices don't. Margin quietly erodes one invoice at a time.
The gap between the price you charge and the price the math supports — that's MarginLab's entire job.
The same logic behind the screen we turn around to clients. Eight passes turn a raw menu, its costs and a month of sales into a defensible price for every item — at the counter and on the apps.
We calibrate a nested-logit demand model so that, at today's prices, it reproduces your actual sales item by item. Items that substitute for each other sit in the same nest, so a move on one shows up in the other.
One café's sales can't reveal price sensitivity on their own, so the engine starts from published research, says so, and shows the whole range — then a short price test (pass vii) measures it.
For one item on its own, microeconomics hands us the answer: the Lerner rule sets the markup over cost from elasticity alone. The engine goes further and solves every price together, because raising one item moves guests to its neighbours.
The optimiser walks the curve and lands on its summit. Too cheap leaves margin behind; too dear sheds more cups than the price makes up for. We find the top.
A textbook optimum from thin data is a liability. So every recommendation is re-run with price sensitivity anywhere from ×0.3 to ×2, and a move that only pays under one assumption is cut back or held.
It's the difference between a model that's right on paper and advice that's safe to act on.
Every item plays a role on your menu. A Traffic Driver earns its keep by pulling people in — it stays cheap on purpose. A Premium Anchor can stretch. We cap each item's allowed movement to its job.
No recommendation ever pushes a magnet out of reach or holds an anchor below its worth.
€4.00 and €3.95 earn nearly the same, but read very differently. We round each optimum to a psychologically tuned ending and keep it clear of barrier prices — the round numbers that make a guest hesitate.
Thresholds are currency-aware: charm endings for euros, clean breaks for rupiah, never tripping a customer over a mental line.
Popularity against profitability sorts the menu into four well-known camps: Stars to protect, Plowhorses to quietly firm up, Puzzles to reposition, and Dogs to fix or cut.
The quadrant tells you not just what to price, but where to spend your attention. Hover any point.
Before we hand over a number, we shake it: price sensitivity, the share of app orders that are new customers, costs and commissions — so you know which figures are load-bearing and which are noise.
Then we measure the biggest one. A two-week switchback test on GoFood (one price this week, the other the next — reversible in seconds) replaces the assumption with what your customers actually did.
The math is calibrated against the street. We plot where your current and recommended prices sit among the cafés around you — unlimited comparison points, not a fixed three — so a move is profit-optimal and defensible to a regular.
The goal is the sweet spot: the most your menu can earn without ever looking expensive.
An illustrative Latte, computed by hand. Watch a single price travel through the passes.
Demand reproduces today's 315 cups; sensitivity starts at ε = −1.50 until a test measures it.
Marginal cost €1.40 → P* = 1.40 · 1.5/0.5 = €4.20.
€3.80 → €4.20 is +€0.40. Apply confidence → a measured +€0.20.
"Core" item, band −3%…+9%. A +5.3% move sits well inside.
€4.00 holds — clean, below the next barrier, no friction.
A "Star". Uplift survives ±20%. Sits mid-pack vs 7 neighbours.
A twenty-cent move, fully defensible — and that's before a single other item is touched.
The audit does this for every item at once, on the counter and on every delivery app — and a price test confirms it before you commit.
A single-item slice of the maths — the classic Lerner rule. Set an item's economics and watch the recommendation, the profit curve and the margin shift in real time.
↳ Headline uplift holds volume; the curve and "volume response" show the elasticity-adjusted reality. The full engine prices every item together, checks your delivery margins and stress-tests price sensitivity — that's the audit.
We're upfront about the edges of the math. These aren't oversights — they're deliberately left out, because each one needs a human who knows your café. That's the conversation.
The model knows your sales, costs and platform fees. What one menu can't reveal is how hard your guests push back on a price rise — that comes from research until we test it.
A coffee often pulls a pastry with it. Firm up the coffee and you might quietly suppress the attached sale. The per-item optimum can't see the whole receipt.
Morning rush and a quiet Tuesday afternoon behave nothing alike. The model averages a month into one figure — it doesn't split the clock or the calendar.
Start with a single conversation, run a full implementation, or keep the menu tuned as costs move. No retainers you can't leave.
Serving independent cafés in Jakarta · Amsterdam · Almere · and remotely worldwide
From your menu to a decision you act on. The whole flow is built so the moment we turn the laptop toward you, the answer is already obvious.
Your café, your market, your currency, your goals.
Items, prices and unit costs go in.
Delivery margins, demand and roles, calibrated.
Where the money is sitting, ranked.
A defensible price for every item.
Tune assumptions, watch outcomes move.
PDF, phone page and workbook — Bahasa & English.
MarginLab grew out of a stubborn spreadsheet — version eleven of a pricing model rebuilt until it was right. A calibrated nested-logit demand model, a delivery audit that reconciles against your own settlement statement, price tests for what a model can't know, role caps, the lot — checked by hundreds of automated tests before it touches a real menu.
The discipline is quantitative; the delivery is human. Independent café owners don't need a dashboard to babysit — they need one clear, defensible move, explained in plain language, on a call. That's the whole company.
Start free: a one-page analysis from your public GoFood and GrabFood prices. The full audit — every item, every platform, a new app price list and a 60-minute walkthrough — is Rp 1,8jt.