Methodology
The Georithm Resilience Method
A site decision you can't defend is a site decision you shouldn't sign. This is the complete method behind the Georithm Score — six weighted inputs, the datasets behind each one, and the arithmetic, written down so a reviewer can check it without our help.
Maintained by Jeremy Agoya, founder of Georithm.
The six weighted inputs
| Input | Weight | Source | Confidence |
|---|---|---|---|
Opportunity Demand scale and growth balanced against how much of the market is already served. | 30% | US Census ACS 5-year estimates; Statistics Canada census releases | Measured |
Competition Competitor density inside the trade-area radius. Lower saturation means more room to trade. | 20% | OpenStreetMap POI features, queried at report time | Measured, coverage-dependent |
Population growth Annualised change in resident population across the most recent comparable periods. | 15% | US Census ACS; Statistics Canada | Measured |
Economic health Spending-power proxy derived from market size, income and growth momentum. | 15% | US Census ACS income tables; Statistics Canada | Modelled proxy |
Climate & hazard risk Exposure to flood, wildfire, earthquake and drought signals for the trade area. | 10% | Open-Meteo climate series and modelled hazard proxies | Modelled proxy |
Infrastructure Density of the built environment and road network supporting trade and access. | 10% | OpenStreetMap; TomTom traffic flow (Professional and above) | Measured |
How to reproduce a score by hand
Each input is normalised to a 0–100 subscore, multiplied by its weight, and summed. There is no hidden term and no post-hoc adjustment.
score = 0.30 x opportunity
+ 0.20 x competition
+ 0.15 x population_growth
+ 0.15 x economic_health
+ 0.10 x hazard_risk
+ 0.10 x infrastructure
# rounded to the nearest integer, 0-100Four standing commitments
Every figure names its source and its vintage
A number without a provenance tag is not evidence. Each input in a Georithm report shows which dataset it came from and which release year, so a reviewer can go to the source and check it.
Modelled inputs are labelled as modelled
Hazard exposure and spending-power are proxies, not insurer-grade or panel-grade measurements. They carry a confidence tag rather than being presented as observed fact.
The weights are published, not proprietary
The six weights above are the whole model. Take the input scores from any report, apply these weights, and you will reproduce the headline score to the same integer.
Stated limits, in the report itself
OpenStreetMap competitor coverage thins out in rural areas, ACS estimates lag by up to two years, and hazard scores are proxies. All three appear in the report, not just here.
See the method applied
The Georgetown, TX sample report shows this exact method on a real market: the map, the competitor list, the input subscores and the GO/NO-GO. No account needed.