Kevin Larretche. Ask about your data

Kevin LARRETCHE.Data Science
× FP&A.

Pricing, promotion and margin studies. The readiness check is free: you get a written verdict on what your data can support before anything is billed.
Based in Hong Kong, from France.

See the studies ↓ Ask about your data
Case studies

Selected studies

Run end to end on public retail data and published in full: the deck, the numbers, the method and the limits.

A shelf of soft-drink cans with their shelf-edge price labels
Price Elasticity
$238k a year, hiding in a price list

Should a supermarket raise prices?

Nine products cleared, four refused by name.
A fashion shop window with a red banner reading sale, up to minus 50 percent
Promotion ROI
2 / 3 of promotions were losing money

Do discounts pay for themselves?

79,340 promotions, each billed for its own discount.
A crowded shopping street under Christmas lights, the season that breaks the budget rule
Forecasting
10% wrong, the way most budgets are built

Still budgeting on last year times growth?

Last year's week is often not the same kind of week. Correct for that and the miss falls to 4.16%.
A wall of nail polish bottles in dozens of shades, arranged by colour
Margin bridge
$300k of the growth was price, not volume

Where did the growth actually come from?

Of $423k in sales growth, barely any was volume. On a generated file, to show the method.
Beyond the published four When the export is missing its price column Readiness check: what your data can support, in writing All studies in detail →
Studies

What each study answers

Reporting tells you what happened. These answer whether the price rise was right, which promotions pay, and why margin really moved.

Usually the first study

Pricing

Where prices can rise, by how much, and what it is worth. Every move shows the volume it should lose next to the volume its margin can afford to lose, both worked out before anything changes.

The pricing study →
Typically the largest leak

Promotions

What each dollar of discount bought back, which depths pay and which burn, and what the displays and leaflets add on top of the price cut. The whole-basket claim gets measured as well, from the receipts.

The promotion study →
When the cause is unclear

Margin bridge

Why margin moved: volume, the blend of products sold, list price, discounts, cost, launches and delistings. The lines add up exactly, and your team can rerun it.

The margin bridge study →
3 of 3
Blind estimates matched the engine, one exactly and all three within 0.04.
79,340
Promotions billed for their own discount, over two years.
3.3m
Baskets read to check whether promotion buyers spent more on the rest of the trip.
0
Hidden steps. Any figure can be traced to the transactions.
How I work

How an engagement runs

Work by project, not by the hour. Nothing is billed until there is a written answer on whether your data can support a decision.

  1. 01

    Readiness check

    One export, and a written verdict on whether your data can support the decision. Nothing is billed for this.

  2. 02

    Go or no-go

    If it cannot, the study is cancelled here, in writing. If it can, one fee and one timetable, agreed before anything starts.

  3. 03

    Ten working days

    The measurement, the checks designed to break it, and the deck, report, workbook and code that you keep. The five steps in full →

Nothing is billed before step one has a verdict, and everything from step five onward is yours, even if we never work together again.
About

I'm Kevin, a data science and FP&A consultant working between France and Hong Kong. I spent my FP&A years at Devialet, owning budgets, closing months and defending forecasts, and I build the statistical models myself, which is why every study ships with its code.

France & Hong Kong English & French Ex-FP&A, Devialet Rennes School of Business Fixed scope, one export Every number checkable