Industrial AI practice  ·  Gulf  ·  Arabic and English Reply within one working day
Core data science (forecasting + decision optimisation)

A mid-market online fashion and homeware retailer, ~14,000 SKUs

Retail / E-Commerce

← All documented proof

The shape of this problem

The same judgment, made differently every time

The problem

Buying and markdown decisions ran on a single blended sales average and merchant intuition, so fast sellers stocked out mid-season while slow movers piled up and cleared at deep, margin-destroying markdowns. Markdowns were set reactively and uniformly, applied late once inventory was already a problem.

What we built

We began with a data and process audit rather than a model, which surfaced that promotions weren't consistently logged and returns weren't netted out - distorting the demand history any forecast would learn from. We cleaned the demand signal first, then built SKU-level forecasts with markdown-timing guidance, prioritising the top-revenue and slowest-moving tiers and piloting on two categories across a full eight-week selling window.

What moved

Measure Before After
Forecast accuracy (WAPE) ~42% error 19% on piloted categories
End-of-season clearance inventory down 27%
Gross margin on piloted categories +3.8 points

† Measured against the prior level of the same measure. The source publishes the size of the movement, not the figure it moved from.

End-of-season clearance inventory

Prior level 0 100 200 down 27%

Measured against the client's own prior process, indexed to 100. The source publishes the size of the movement, not the absolute figure it moved from.

Figures are drawn from the practice's own delivery records for the engagement named, measured against the process that preceded it. They have not been through third-party audit, and none is presented as an average across clients.

Our own sales history was lying to us because of how we logged promotions and returns - fixing that was unglamorous, and it's where most of the accuracy gain came from.

Head of Merchandising

Service line
Forecasting & Planning AI
The finding matters more than the number here. Both cases repaired data definitions before modelling, which is the practice's whole thesis in miniature.

Start here

Which of the four is yours?

Tell us the documents and the monthly volume and we will send the two closest records, with the proof behind each and the honest note where the match is partial.

You get a reply within one working day, from the engineer who would do the work - not a sales sequence.