Case study · Retail AI

Mealyn

A retail technology experience that turns household preferences, budget and food choices into a personalized grocery list — delivered through web software and an in-store self-service kiosk.

CategoryRetail tech / foodtech
TouchpointKiosk + web
LogicPersonalization + recipes
Customer using Mealyn kiosk
30stargeted in-store decision flow
360°household preferences + budget context
WEBrecipes and product experience
PRINTphysical output at the kiosk

From a vague question to a concrete basket

“What should I buy?” sounds simple until software has to answer it for a real household. Preferences, number of people, budget, dietary constraints, product availability, offers and recipe logic all interact.

The engineering challenge is to convert that complexity into a short sequence a shopper can understand while standing in a supermarket.

Design for a physical environment

A kiosk is not a desktop website on a large screen. People stand at a distance, touch instead of click, expect immediate feedback and may leave after a few seconds. Targets have to be obvious, flows short and the product must recover cleanly when a session is abandoned.

Personalization without interrogation

The question flow needs enough information to build a useful result without feeling like a survey. Each question therefore has to justify its place in the decision model.

A useful result, not an AI demo

The output is concrete: a grocery list and associated recipes that can be adjusted, printed or used later. AI and recommendation logic are successful only when the shopper gets a more practical next action.

The best model output is the one that becomes a better shopping decision.

Retail operations matter too

The supermarket side can use the platform to think about promotion, product movement and customer experience. That makes Mealyn both a consumer interaction and a retail operating product.

Continuous product development

Real kiosk usage creates the feedback loop: where customers pause, which questions confuse them, which recommendations they accept and whether the final list feels credible. Those signals drive the next iteration.

Experience architecture

A physical product powered by software decisions.

Kiosk session
householdbudgetpreferencesdiet
Recommendation layer
offersproductsrecipesconstraints
Output
shopping listrecipesreceiptemail/web
Retail layer
availabilitypromotionpilot analytics
recommendation.ts
const plan = await engine.buildList({ people: household.size, budget: session.budget, preferences: session.preferences, exclusions: session.dietaryConstraints, products: store.availableProducts, offers: store.currentOffers }) return output.compose({ list: plan.items, recipes: await recipes.match(plan), route: store.sortByAisle(plan.items), printable: true })