The chain's board had set a clear ambition: an AI answer for the loyalty programme, in-store pickup and the returns operation, all three at once. What arrived instead was noise. Ideas came in from every department, each convinced its problem was the priority, with no shared method for weighing a loyalty idea against a returns idea or a pickup idea.

The data underneath those ideas was just as scattered. Information relevant to loyalty, pickup and returns sat in four separate systems, so even a well-formed idea could not easily be checked against the data it would need. Two early vendor demos had already stalled for exactly this reason: nobody could define, in terms the data could answer, what problem the demo was actually meant to solve.

Without a shared evaluation method, the loudest department tended to win the argument rather than the strongest idea, and the board's ambition for loyalty, pickup and returns stayed exactly that, an ambition, while two stalled vendor demos left staff wary of the next pitch.

We designed the discovery programme to replace department-by-department pitching with one shared process. Nine cross-functional workshops brought merchandising, loyalty, stores and operations into the same room, so a loyalty idea and a returns idea were finally being compared on the same terms.

Alongside the workshops, data walkthroughs of the four core systems established what the underlying data could actually support, closing the gap that had stalled the two earlier vendor demos. Opportunity sizing against two years of transaction history then let every candidate use case be checked against real evidence rather than assumption.

Each candidate use case was scored on value, feasibility and data readiness using the same criteria, then shortlisted into costed business cases with a recommended delivery order, so the executive committee could fund a sequence rather than argue over which department went first.