UC-051
Counterfactual Site Selection
Ranks planned store openings by predicted revenue, cannibalization risk and permit cycle time before capital is committed.
TBDBuild Duration
XXXXIndicative ROI
XXXXBuild Cost
XXXXValue
The Challenge
Significant capex per store and hundreds of openings a year mean weak site selection misallocates hundreds of millions. Permit filing is slow and site calls rest on judgment, not counterfactual evidence.
How It Works
- Embeds each site in local demographic, competitive and demand context with a graph model.
- Predicts revenue and own-store cannibalization using doubly-robust causal estimation.
- Automates municipal permit filing through a generative agent.
What It Removes
- Site calls made on local judgment alone
- New stores that quietly cannibalize existing ones
- Permit filings that sit in a manual queue
Input Data RequirementsCandidate site list, local demographic data, competitor locations, demand data, existing store network, municipal permit requirements
Output FormatRanked site shortlist with predicted revenue and cannibalization, automated permit filings, capex allocation view
“We can see which of the next hundred sites actually earn their capex, and the permits are already moving.”