Frequently asked questions

Yes. Every forecast of a location that hasn't opened yet is an estimate, and some land outside the expected range. That is exactly why IRIS gives you the forecast with its prediction interval, never a bare number, validates out-of-sample on your own openings before you rely on it, and is upfront about where the model is weaker. A tool that never admitted a miss would be the one to distrust. See how we validate.

A general-purpose language model can summarise what is already known about an area, but it does not forecast revenue for your specific store, and it cannot tell you how confident to be. IRIS is a statistical and machine-learning model calibrated to your own estate and more than 100 local signals, then validated against real openings. Different job, different tool. More on what goes into a forecast.

No. IRIS is built for the person who owns the expansion decision: point-and-click maps, scenarios and reports, no GIS or data-science background required. Your data team gets a documented API and full data export, but nobody has to open a laptop full of code to get an answer.

Yes, and yes. Alongside the point-and-click platform, IRIS has a documented API and an MCP (Model Context Protocol) server, so your data scientists can drive everything IRIS does, running scenarios and opening, closing or relocating locations, from their own assistants, agents and scripts. The built-in assistant inside the platform is private by design: it works through IRIS's own tools rather than your raw data, and only ever sees the result of a query, a score, a forecast, an impact summary, never your underlying store, customer or transaction records. So sensitive data is never handed to a language model, and each forecast still shows which signals mattered most, so you can explain a score to a board rather than just quote it.

Whenever the prediction interval is wide or the model is working with thin data: a brand-new market with no comparable stores, a very small store count, or an area with sparse public data. IRIS flags these cases rather than hiding them behind a confident-looking number, so you know when to lean on local judgement. See where the model is weakest.

Three answers, because they are genuinely different things. Our standardised data layer is built and current for every European country except Russia, Belarus and Ukraine, plus Turkey, so IRIS Market runs there today at the published price. We have modelled sites in 26 countries so far: Ireland, Norway, Sweden, Finland, Denmark, Estonia, Latvia, Lithuania, Poland, Germany, the Netherlands, Belgium, Luxembourg, France, Switzerland, Portugal, Spain, Croatia, Hungary, Romania, Bulgaria and Greece, and outside Europe in Morocco, Saudi Arabia, Taiwan and Mexico. Anywhere else, we build the data layer for that market first, and the Validation Sprint scopes what that takes before you commit to anything.

Your data stays yours. The Big Data Company is a European company and IRIS is subject to European data-protection law (GDPR). You keep full export and API access to everything, and your own store performance is used to calibrate your models, not sold on.

The licence is priced per market, per year, on the revenue your stores in that country take at the till, because that is the money the decision is made against. Above €25M of revenue in a market it is at most around 0.1% of it, and under 0.01% on the largest estates. Below that a floor applies: €16,000 on IRIS and €9,000 on IRIS Studio. Seats are included and we do not meter scenarios. You can work out your own figure on the plans page without talking to anyone. IRIS Market is €6,000 a year per market and scores locations on our own standardised data, though it will not forecast revenue, because we do not have your sales data. To forecast on your own data, start with the Validation Sprint at €4,000 for the Kick Start, €12,000 all-in, which is the first build sprint and credits in full if you go ahead. On the IRIS tier the accuracy is then committed in writing, and if the live model misses it your first-year licence for that market is refunded, where that market has 50 or more comparable mature sites.

You do. It trains your model, not ours and not a competitor's, unless you opt in to collective learning. This is contractual, not a reassurance, and you can ask us for the clause. EU-hosted, no US mobile panel, LIA and DPIA on file and readable.

IRIS Market is self-serve by design, which is how it stays at €6,000 a year. Bugs we fix, that's our job, not support. When you want someone to study your catchment with you, that's Studio, and it's priced. Same principle as the revenue number: when IRIS Market can't answer something honestly, it says so rather than guessing.

No, and that's deliberate. We go deep on the model and the calibration, and stop there. Rolling a new method through an organisation, training teams, changing how decisions get made, is real work that belongs to specialists. Where you need it, we partner with implementation firms who do it for a living: we own the forecast and its honesty, they own the rollout.

None yet: the paid Kick Start is new, and the first hasn't run. From that first check onward we publish two numbers here, how many we've run and how many estates we declined, and we keep publishing them when the second is inconvenient.