Method
How platforms get tested
The short version: nothing gets recommended here that hasn't been built on, and pricing is modelled at volumes agencies actually reach.
Hands on the tool
Every platform covered is one I've either used on a paying client project or built a genuine test workload on. "Genuine" means real records with real edge cases — duplicates, missing fields, rate limits — not a five-record happy path.
Pricing modelled at volume
Vendors quote the tier that makes them look cheapest. Comparisons here model cost at the volumes that actually matter, and state the assumptions. Where two tools count usage differently — a Zapier task is not a Make operation is not an n8n execution — the difference is explained rather than papered over.
Dates on everything
Every price carries the date it was checked. Software behaviour carries a version where it matters: "tested on n8n 1.x, 2026-08-28" is useful, "updated recently" is not.
What gets said about weaknesses
Every tool profile names what the tool is bad at. A review that can't say what a product does poorly isn't a review.
Corrections
Vendors change pricing and features constantly. If something here is out of date, tell me and it gets fixed with the date updated. A page's reviewed date only moves after a real re-check.