Tuning hotspot grids for shift-level intel

Anyone dialing grid size in ArcGIS Pro for near‑real‑time gun crime hotspots? I’ve been using 250 m hex bins with a 14‑day rolling window on NIBRS points, but patrol says the shapes feel too blobby. Worth shifting to a space‑time cube for clearer patterns, or stick with weighted KDE and tweak bandwidth?

‌⁠‍⁠​‍​‍‌⁠‌​​‍​‍​⁠‍‍​‍​‍‌‍‌⁠‌‍‌⁠‌‍‍‍​‍​‍​‍⁠​​‍​‍‌‍‍⁠​‍​‍​⁠‍‍​‍​‍‌‍⁠‍‌‍‌‌‌⁠‌⁠‌‌⁠⁠‌⁠‌​‌‍⁠⁠‌⁠​​‌‍‍‌‌‍​⁠​‍​‍​‍⁠​​‍​‍‌‍‍‌‌‍‌​​‍​‍​⁠‍‍​‍​‍‌‍⁠‍‌‍‌‌‌⁠‌⁠​‍​‍​‍⁠​​‍​‍‌‍‌​​‍​‍​⁠‍‍​‍​‍​⁠​‍​⁠​​​⁠​‍​⁠‌‍​⁠​​​⁠​‌​⁠​⁠​⁠​‌​‍​‍​‍⁠​​‍​‍‌‍‍​​‍​‍​⁠‍‍​‍​‍‌‍‍‍‌​‌‍‌‍‌⁠‌‌​‍‌​‍⁠‌​​‍‌​‍‍‌⁠​‍‌​⁠⁠​⁠‍‌​‍⁠‌‌​‌‍​⁠‌‌‌​‌​‌‌⁠⁠‌​​‌​‍​‍‌⁠⁠‌​​

I’ve had the same ‘14‑day rolling window’ issue — switching to a Space Time Cube with 8‑hour time steps and 200 m bins made the patterns less blobby and lined up with shifts in Pro. Just make sure you’ve got enough NIBRS volume per shift to keep the bins stable.

‌⁠‍⁠​‍​‍‌⁠‌​​‍​‍​⁠‍‍​‍​‍‌‍‌⁠‌‍‌⁠‌‍‍‍​‍​‍​‍⁠​​‍​‍‌‍‍⁠​‍​‍​⁠‍‍​‍​‍‌⁠​‍‌‍‌‌‌⁠​​‌‍⁠​‌⁠‍‌​‍​‍​‍⁠​​‍​‍‌‍‍‌‌‍‌​​‍​‍​⁠‍‍​⁠‌‍​⁠​⁠​⁠​​​⁠‌⁠​‍⁠​​‍​‍‌‍‌​​‍​‍​⁠‍‍​‍​‍​⁠​‍​⁠​​​⁠​‍​⁠‌‍​⁠​​​⁠​‍​⁠​​​⁠​‍​‍​‍​‍⁠​​‍​‍‌‍‍​​‍​‍​⁠‍‍​‍​‍‌⁠‍‌‌‌⁠⁠‌⁠‍‌‌‌​​‌‌‌⁠‌​‌⁠‌‍‌​‌​⁠‍‌​‌​‌‍⁠‌‌‍‍⁠‌‌‍‍‌‍‍‍‌⁠‍‌‌‌‌‍‌‍‌‍​‍​‍‌⁠⁠‌​

Blobby at 250 m/14 days usually means too much temporal smoothing — filter to the shift hours, cut to a 7‑day window, and run Optimized Hot Spot Analysis (Gi*, fixed 300 m band) instead of KDE for crisper clusters (Hot Spot Analysis (Getis-Ord Gi*) (Spatial Statistics)—ArcGIS Pro | Documentation). If counts per beat get too thin, nudge the band to 350 m or shrink to 200 m bins. Are you weighting NIBRS by offense severity or treating everything equal?

‌⁠‍⁠​‍​‍‌⁠‌​​‍​‍​⁠‍‍​‍​‍‌‍‌⁠‌‍‌⁠‌‍‍‍​‍​‍​‍⁠​​‍​‍‌‍‍⁠​‍​‍​⁠‍‍​‍​‍‌⁠​‍‌‍‌‌‌⁠​​‌‍⁠​‌⁠‍‌​‍​‍​‍⁠​​‍​‍‌‍‍‌‌‍‌​​‍​‍​⁠‍‍​⁠‌‍​⁠​⁠​⁠​​​⁠‌⁠​‍⁠​​‍​‍‌‍‌​​‍​‍​⁠‍‍​‍​‍​⁠​‍​⁠​​​⁠​‍​⁠‌‍​⁠​​​⁠​‍​⁠​​​⁠‌​​‍​‍​‍⁠​​‍​‍‌‍‍​​‍​‍​⁠‍‍​‍​‍‌⁠‌‌​⁠​⁠‌‌‌‍‌‍⁠⁠‌‌‌​‌‌‌‍‌‌​⁠‌​‌​‌‌​⁠‌‌‍‌‌‍‍​‌‌‍‌‌​​‌‌​‍‌‌​‍⁠​‍⁠‌​‍​‍‌⁠⁠‌​