Making allocation recommendations clear enough to trust

Global merchandisers balanced stock across thousands of stores and warehouses. The work moved between SAP, large spreadsheets and local rules held in people's heads. Comparing options was slow, exceptions were difficult to see, and experienced planners carried much of the system's operational memory.

Allocation predictions already existed, but they arrived as opaque answers. Planners could not see why a market received a quantity, compare the signal with their own knowledge or record an intentional exception. The team did not need another dashboard. It needed a decision surface.

1. 01

   Clarity

   Show the recommendation beside demand, historical variation and the exception that needs attention.

2. 02

   Trust

   Show the model's context so planners can inspect the basis of a suggestion before acting. A confidence score was designed for this but did not ship.

3. 03

   Agency

   Keep override, simulation and approval explicit. The system proposes; the planner remains accountable.

Tension

The AI pre-filled each value from a prediction matrix. Approving it blind would make planners rubber stamps; rejecting it blind would waste the prediction.

Choice

Each default sat next to the four things a planner checks: major shipment orders, the previous two years of data, aggregate costs, and any decisions agreed for that in-shop week (the week stock is due in store). Nothing went through without manual approval.

Trade-off

Approving every default by hand was tedious, and the MVP accepted that on purpose. The approvals built the trust and the record needed before the system could take over routine decisions.

Tension

Each recommendation drew on two years of history, plus live costs for ordering, shipping and raw materials. It couldn't know what marketing and merchandisers had agreed between themselves: a larger order, or ramping up an existing line.

Choice

Planners could accept or override every recommendation, and the existing checks stayed in place.

Trade-off

Because the human checks stayed, the tool cut file handling more than decision time. Speed was given up to keep planners accountable.

Considered

A larger, less dense layout closer to Fabric's consumer defaults.

Chosen

Compact rows and enterprise table patterns. We tested both with planners, and people used to sifting through 20 Excel files and thousands of rows preferred the dense one.

Trade-off

Smaller type and targets than Fabric's defaults, a deliberate accessibility compromise for a trained internal audience. Fabric was built for consumer screens, so the enterprise patterns had to be negotiated with the design system team.

Make the machine legible, keep expert judgement active and give disagreement a safe path through the system.