AppFolio sync at scale

Why Western Slope's first data load choked tonight, and the fix: a one-time import for new customers, a slower everyday check for existing ones.

2026-09-13 · Fable for Fede · 5-stage import shipped, JP&Co arming in progress, Western Slope re-import in progress · 2 decisions open

The verdict, in plain English

We built one system to check on customers who already have data. Asking it to load a brand-new customer's whole portfolio from zero burned through 25,000 requests in three hours and still left 54 of 69 buildings unfinished.

The fix, now built: a one-time import that pulls a new customer's data in five ordered stages, and a slower everyday sync that calls AppFolio once per customer instead of once per property — three speed classes, one shared budget per account, every job's portfolio-wide call its own dark switch.

The math: a fully-loaded property costs ~6,200 AppFolio calls a day today. For Western Slope's 69 properties, that's the ~429,000/day the old design implied; the new design does the same job in ~6,200/day total — about 69x fewer, whatever the portfolio size.

Tonight's arming (budget, occupancies, guest_cards) is proving out on JP&Co's real account, one switch at a time. Western Slope's own re-import is running now. Nothing goes live at a real customer until it's proven dark, then at Willows, then Fede turns it on.

PropFlow Docs