Last month I was standing in a strip center lot in Round Rock at 5:30 p.m. comparing our 10-minute drive-time to the old 3-mile ring; the river crossing and a nasty left turn cut the effective trade area in half, our cannibalization model flipped from +8% to -3%, and we killed the site. Anyone here leaning on Network Analyst or H3 grids for quick site screens — and what travel-time sources are you trusting lately?
Quick example: we started snapping candidate sites to the actual driveway and modeling right-in/right-out with a U-turn penalty, which chopped the ‘10-minute’ isochrone in half on the PM peak. If you’re in Network Analyst, time-dependent speeds make that stick: https://pro.arcgis.com/en/pro-app/latest/help/analysis/networks/understand-time-dependent-networks.htm. H3 is great for triage, but I still sanity-check a couple routes in Google Maps.
I keep a pre-baked H3 (res 9) grid of 10-minute drive times by daypart and flag cells that shift >20% at “5:30 p.m.” — bridges and school zones in Round Rock light up, so sites near those get auto-yellow before I even model turns. Then I’ll borrow @ptaylo64’s driveway detail if it survives; only caveat, H3 glosses over gated subdivisions unless you burn access nodes.
, I’ve had better luck baking in toll aversion and a 30s “first-signal” delay at 5:30 — a bunch of folks won’t take SH 130, and when the “river crossing” clogs it turns a keep into a kill fast. I run it in Valhalla so I can flip profiles by daypart; skip the driveway signal and your polygon looks way too optimistic. Small caveat: the toll penalty can over-shrink in higher-income pockets, so I back it off there.
But i split catchments into inbound vs outbound and add a 45–90s egress penalty for unsignalized left exits; probe traces showed sites that looked fine on the way in but were crab traps on the way out at 5:30… Building on @chen54, I also flag school-pickup windows and temporary turn bans so the “river crossing” effect doesn’t surprise you, and I bake those costs in with Valhalla’s custom costing (https://valhalla.readthedocs.io). If you don’t want to go that far, even a simple signal-per-mile factor will keep a borderline site from fooling you.
.