RetailIQ Site Report

A team opens a Site Report to answer one question: should we open a store here? The report drew on 50+ KPIs and the outputs of several models, but it mostly said “here is the data” rather than “here is what it means for your decision”. Clients, Sales and Product leadership kept raising the same issues: what should come first, which numbers deserve a stronger signal, and how the same information should be shown. The report was also the moment a new user first understood RetailIQ, and it arrived only after a long onboarding.

Note: 50 numbers. One real question: open here or not?

The moments of uncertainty

• ?Where do I start? The report opened on raw data instead of a conclusion.
• ?Which of these 50+ KPIs actually matter for my category?
• ?Is this number good or bad, and compared to what?
• ?Where is this happening? A percentage doesn’t tell me which stores or streets.

Helping people find their bearings

I restructured the report around how people read a decision: Data → Signal → Evidence → Decision. Instead of showing a footfall or cannibalisation number on its own, the report says whether it is high or low, shows the trend or the place it comes from, and keeps the underlying data one step away. I worked with the data and model teams to understand what each output actually meant, so the interface represented it accurately.

From an AI highlight to the evidence

The Site Report uses AI-generated highlights and model outputs to surface notable findings from a much larger body of location data. I didn’t build the models. My focus was the decision layer around them: what comes first, how each signal is phrased and shown, and how people move from a highlight to the evidence behind it.

1. AI highlight
2. Decision signal
3. Visual evidence
4. Detailed data

Note: AI points at what matters. The numbers stay one click away.

A highlight is a starting point, not a verdict. Every one can be traced to the data that supports it, so people can check it before they act on it.

What I chose — and why

1. 01

   Lead with the signal

   A score, a recommendation and AI-generated highlights such as “high footfall” or “high cannibalisation” answer the question on the first screen.

2. 02

   Pick the right representation

   For each KPI: a meter, an indicator, a trend, a graph or a map? Cannibalisation, for example, works better as a map of where it happens than as a percentage.

3. 03

   Six layers, one question each

   Detailed data sits under each layer, so experts can dig in and everyone else isn’t overwhelmed.

4. 04

   A report that teaches the product

   The report doubles as an entry point to RetailIQ, which led to the idea of letting people experience a report before asking for all their business context.

Product

Reframed the report from a data presentation into a decision experience: signal first, evidence next, raw data on demand.

Capability

AI-generated highlights, decision signals and six structured layers give hierarchy to 50+ KPIs, so people can move from a conclusion to its evidence.

Input

The direction was shaped by feedback from clients, Sales and Product leadership on what the old report made hard.

Product scale

17k+ reports generated across 114k+ evaluated locations. These numbers describe how much the product is used, not results of the redesign alone.

A report is not a dashboard. Dashboards are monitored; reports are read once, carefully, and then defended in a meeting. Designing for that moment changed every decision about what goes first.