Northwind moves three hundred thousand parcels a day across the British Isles. Until last March, we moved them with four observability vendors, two deploy systems, and a custom pipeline gluing the whole thing together. The platform team spent two thirds of every quarter keeping the platform alive.
The thing that changed wasn't the number of tools. It was the surface area. When the dispatch service started missing its SLO at 3am, on-call had to log into three dashboards to see the same request. Stradiva gave us one timeline. The first night with Stradiva, we caught a regression in fifteen seconds that would have taken us an hour to triangulate.
We migrated in phases — observability first, then deploys, then the data layer last. Each phase paid for the next. By the end of the first month, two engineers were back on product work. By the end of the third, we'd shut down the internal platform team entirely and let those engineers pick where they wanted to land.
The honest measure of a platform isn't its dashboards. It's whether your senior engineers stop looking at them.
— Iris Chen, Staff Engineer, Northwind