Know which promotions pay off before you run them.

skuel predicts how your shoppers will respond to any price or offer, so you launch the winners and skip the rest.

skuel

Plan in plain words

Ask for the plan you want.

Tell skuel your goal and your limits: how many products, how deep, what budget. It runs thousands of synthetic experiments and hands back the plan that earns the most.

A planner asks skuel for a 12-product plan at most 30% off; skuel replies with the best plan, net profit $4,959

From nearly 30,000 products in the skuel demo: 12 offers, all profitable.

Test before you launch

Find the offers that lose money.

Already have a plan? Upload it. skuel predicts the extra units, the margin given away, and the sales pulled from your other products, for every offer.

Simulation of a 500-promotion plan showing that 449 promotions lose money

A 500-offer plan uploaded to the same demo: 449 offers cost more than they earn.

Why it matters

Price is your strongest lever. It's also the one you can't safely test live.

TrustShoppers notice when they see a different price than someone else.

TimeA clean live test takes weeks, so most ideas never get tried.

MarginUntested discounts go to shoppers who would have paid full price.

How it works · 1

It starts with how your shoppers behave.

skuel follows each product's shoppers: who hesitated, who left items in the cart, who bought a competitor instead. That shows where price matters, and where it doesn't.

Session flow for one product: engaged, added to cart, abandoned, and bought a competitor

How it works · 2

Thousands of experiments. None of them on your customers.

From that behavior, skuel runs thousands of synthetic experiments in the background to find the promotions that pay. You only see the few worth launching.

0 experiments tested0 worth running

Typical scenario toolsskuel

Who does the searchingYou pick each scenario to testTests thousands on its own

What it learns fromLast year's sales and promotionsHow shoppers behaved this week

A brand-new offerNo history, so a guessPredicted from shopper behavior

95%

of the accuracy of models trained on years of sales data, without training on yours.

36Mreal shopping events tested, from a major European e-commerce platform

73%less error than GPT-5.5 on its own

75%less error than Claude Fable 5 on its own

See the research Method, results, and whitepaper

Built by ex-industry AI researchers.

Vijay Kamble

Co-founder and CEO

Redesigned how Amazon sources deals for Prime Day. Associate Professor at the University of Illinois Chicago. PhD, UC Berkeley.

Varun Gupta

Co-founder and CTO

Builds optimization and machine learning for large marketplaces. Professor at the University of Utah. Formerly Microsoft, Google, Bell Labs. PhD, Carnegie Mellon.

See your next promotion before your customers do.

We're working with a few retail and DTC brands as design partners. Tell us a little about yours.

Works with the session data you already collect. Anonymized, and it can run in your own cloud.