Waste: Real-time markdown pricing and food-waste platform

Sell more. Waste less.

We engineer Waste, a real-time markdown pricing platform that gives store associates the best reduced price for a perishable item from a single barcode scan, so retailers sell more and waste less.

Retail / Grocery and convenience
Data engineering
Software development
Project

10%+total waste reduction

6%+sales recovery

Up to 90%sell-through

<200 mstarget response per scan

The Challenge

What stood in the way.

Retailers were losing margin on perishables by discounting too much, too early, or too little, too late. Markdowns were decided by judgement on the shop floor, with many factors to weigh for every item and no time to weigh them.

Inconsistent pricing

Judgement-based markdowns produced different prices between associates and between stores.

Margin lost both ways

Discounts that were too deep or too early gave away margin on stock that would have sold. Discounts that were too shallow or too late left unsold stock to be thrown away.

Price needed at the shelf

Expiry, time, store and volume all matter for each item, and batch reports arrive too late to help an associate holding the product.

Different rules per retailer

Each retailer has its own pricing rules and data formats, so custom builds per retailer were slow and costly.

Our Solution

How we answered it.

We built the platform in three parts: an offline side that prepares answers, an online side that serves them instantly, and a control plane for changing pricing rules. Every pricing factor combination is calculated overnight

ProblemComplex pricing maths had to fit inside a 200 ms response limit.
Our answerWe pre-compute every factor combination offline and serve each price by lookup.
ProblemHeavy data loads could not be allowed to slow live store traffic.
Our answerWe separated the processing and serving databases, with an optional Redis cache in front of the API.
ProblemData had to refresh without interrupting stores.
Our answerNew data loads into a copy and the tables are swapped in one atomic step, giving zero downtime during refresh.
ProblemEvery retailer sends files in a different format, some with poor-quality rows.
Our answerEverything loads as text first, then is typed and validated. Bad rows go to a review table and a load fails only past an error threshold.

How It Works

The parts that made the difference.

Pre-computed pricing engine

An overnight T-SQL process calculates every pricing factor combination into one lookup table, so live pricing never waits on heavy maths.

Pricing API

An ASP.NET Core REST API reads the barcode, validates the item and store, applies the retailer's rules and returns the price with a rule trace, with optional Redis caching.

Rules as configuration

Per-retailer JSON workflows on Microsoft RulesEngine support set, adjust, rounding and limit actions. A deploy service pushes saved rules to each retailer automatically.

Store associate app

A .NET MAUI app with SSO, camera scanning, price display, authorised overrides, date-check reminders, push notifications and electronic shelf label integration.

Results

What changed for the client.

Retailers get a consistent, rule-based markdown for every scanned item, and pricing teams can adjust strategy in hours rather than sprints. Published results show waste falling and sales recovering, with pilots going live in weeks.