Where to Open the Next Store? Location Selection with Mobility Data
A specialty coffee shop chain has 140 outlets and wants to open 25 more next year. Each opening costs €350,000 and 20% of the latest openings closed within 3 years. The expansion team has 300 premises available for rent and needs to estimate first-year sales for each one.
Where the money comes from
What to do now
- 1Include closed stores in the history.
- 2Estimate sales and payback for the 300 candidate premises.
- 3Subtract cannibalisation and rent for each candidate.
For your committee
One page with the decision, the impact, the risks, the steps and the KPI to track.
The print dialog opens: choose "Save as PDF".
Watch out: more volume is not always more profit
Opening where the highest sales are forecast without subtracting cannibalisation of nearby own stores or the more expensive rent: more gross sales, less profit for the network.
KPI for the committee
Payback of openings and % of closures within 3 years.
Detail for your data team
Proposed technique: Random Forest
Why: With only 140 stores, a very complex model would overfit. Random Forest is robust with little data, captures non-linear relationships (a terrace matters more near offices) without much tuning, and the spread between trees gives an idea of the uncertainty of each forecast.
Cost of getting it wrong: A bad location means losing €350,000 of investment and years of rent; discarding a good location leaves space for the competition.
Data needed: First-year sales of the 140 current outlets.; Anonymised mobility data: pedestrians per hour on each street, visitor profile.; Surroundings: offices, universities, stations, competitors within 300 m, average neighbourhood income.; Premises characteristics: square metres, terrace, visibility, corner.