September 24, 2026 · Tupll
How to Compare Candidate Locations When Every Site Looks Good on Paper
The hardest site decisions are not between a good site and a bad one. They are between three sites that all look good, each with a champion, each with a broker package full of favorable numbers, each defensible in a meeting. When everything looks good on paper, the paper is the problem.
Why good-looking candidates are hard to separate
Broker packages and standard reports are built from the same public data, cut to flatter the site. Every package shows population, income, traffic counts, and growth. And by those measures, most shortlisted sites genuinely are fine; that is how they made the shortlist.
The differences that matter live below that layer: the composition of demand rather than its volume, access friction that traffic counts hide, competitive interception, and overlap with your own existing stores. None of it shows up when each site is presented as its own brochure.
Rank, don't audition
The structural fix is to stop evaluating sites one at a time and start ranking them on a single model. Same variables, same weights, same method, applied to every candidate, with the weights learned from your own locations' performance rather than industry averages.
Ranking changes the meeting. Instead of three champions defending three brochures, the room looks at one list: this site scores 94, this one 81, this one 63, and here is which variables drove each score. Disagreement becomes specific ("why does access weigh so heavily?") instead of tribal ("I just like the Northside site").
Expect the ranking to surprise you
If a model never contradicts your intuition, it is not adding information. The valuable moments are the reversals: the site everyone favored ranking second, because its impressive surroundings are the wrong composition; the overlooked site ranking first, because it concentrates exactly the customers your best stores share.
In one of our engagements, two candidate sites in the same metro looked like a blowout on eyeball metrics: one had roughly six times the households and five times the businesses nearby. The model ranked the smaller market first by a wide margin, projecting about 32 percent more first-year revenue, because it held the buyers that brand actually converts. The obvious pick would have been the expensive one.
A practical protocol
Three habits make candidate comparison rigorous. Score all candidates before anyone presents a favorite, so the ranking anchors the discussion rather than rebutting it. Require every score to be explainable, so champions can argue with the weights instead of the conclusion. And record the ranking with the decision, so when year-one results arrive you can check the model against reality and make the next comparison sharper.
Paper makes sites look alike. Evidence makes them come apart.