Guides
Guides
Short, honest explainers for the people who own an expansion decision, not just the data team. No jargon for its own sake, and no claims we can't validate.
- Prediction intervals Why won't IRIS just give me one number? IRIS shows a revenue forecast with a prediction interval, not a single figure, because a lone number hides the uncertainty you need to make a defensible site decision. Read · 3 min
- Explainability Which signals actually drove this location's forecast? A revenue forecast is only trustworthy if you can see why it landed where it did. Here is how feature importance and interaction effects explain a location score, signal by signal. Read · 3 min
- Validation Out-of-sample validation vs overfitting: how to tell a forecast will actually predict A model that fits your existing stores perfectly can still be useless on the next one. Here is the difference between overfitting and out-of-sample validation, and how to check a vendor's accuracy claim. Read · 3 min
- Method · Step 1 How IRIS draws a catchment Why a catchment is not a radius or a postcode. How IRIS builds it from travel time on foot, by bike and by car, weights it by distance, and lets real barriers cut it off at the door. Read · 3 min
- Method · Step 2 Reading the demand behind a site How IRIS turns 100+ local signals into seven demand drivers you can read at a glance, and why the reason people are near a site, plus when and how long, matters as much as how many. Read · 3 min
- Method · Step 3 Why we validate out-of-sample A forecast means nothing until it is tested against stores whose real revenue you already know. How out-of-sample validation works, why most tools skip it, and why some error is irreducible. Read · 3 min
- Method · Step 4 What the 80% interval means How to read the 80% interval plainly: four times in five, actual revenue lands in the range. Why the range is the answer rather than a hedge, and why site selection carries fundamental uncertainty. Read · 3 min
- Method · Step 5 The auditable forecast brief The auditable brief: every forecast arrives as a brief you can open, question, and re-run. See the value, its interval, and the drivers behind it, and defend the number where capital is committed. Read · 3 min
- Demand drivers The demand drivers, and what feeds each The seven demand drivers IRIS uses to read the demand behind a site: what a demand driver is, the input data that could feed each one, and how much each tends to matter. A practical decomposition, refined over time, not a fixed law. Read · 6 min
- Transparency What the data layer actually costs We publish our own data-cost basis. What the standardised European data layer inside every IRIS licence would cost to source yourself, per data type, at H3 resolution 10 (~120 m hexes) and full-market coverage, with the suppliers and the sources named. Read · 5 min
- Choosing a tool Questions to ask any site-selection vendor A short due-diligence checklist for choosing a location-forecasting tool: can you see the reasoning, is it checked on your own stores, does it tell you how sure it is, can a person overrule it, and do you keep your data and your model. Read · 4 min
Ready to see IRIS on your own market?
Book demo