StreetSpring Survivability Datasets 2026: projected two-year survival of 110 business types by neighborhood across 24 U.S. metros

Projected chance, in percent, that a specific business type lasts two or more years at a specific place, from a model trained on 570,000+ observed business outcomes and 100+ location factors in 24 U.S. metros. Every score column is the projected two-year chance on the same calibrated scale the StreetSpring rankings pages use; it is never the model's raw output. One national file (metro by business type, 2,595 rows) and 24 metro files (neighborhood by business type), 110 business types per file, 380,015 rows in total. Each metro file's scores average across the price points StreetSpring scores separately in its tool (most types have 5). README.md carries the column dictionary and recipes for reading the files; CHANGELOG.md lists what changed in each version. Version 2026.09.25: rankings follow p90_survivability_score, the chance of lasting two or more years at a good address (avg_survivability_score is the typical address). New columns neighborhood_overall_score, neighborhood_overall_avg_score, neighborhood_overall_max_score, neighborhood_overall_min_score, business_types_ranked_in_neighborhood, neighborhood_overall_rank and total_ranked_neighborhoods_overall. Two location-count columns are renamed scored_locations_in_neighborhood and scored_locations_in_city, values unchanged. 1,317 neighborhoods added from city and county boundary files; non-places removed from the public files. The same files, with worked answers, are on the dataset pages at streetspring.com/resources/datasets. Source repository: github.com/StreetSpringAI/streetspring-survivability-datasets. Methodology: streetspring.com/resources/tutorial/aeo-methodology. License: CC BY 4.0.