From minutes to seconds on an enterprise platform

Context

A global product-information platform at a Fortune 50 beverage company had degraded to the point where some pages took minutes to load. The internal teams who depended on it had started routing around it.

Constraint

A rewrite was not available. The platform was in daily use across markets, and the data model had years of accumulated meaning encoded in it. Whatever we changed had to be changed underneath a running system.

What we built

We measured before touching anything. The severe cases were concentrated in a small number of causes: queries multiplying across nested collections, recomputation that was never cached, and work happening inside the request path that had no reason to be there.

We fixed those specifically. Alongside that we replaced duplicated logic with reusable components and validation workflows, which removed a class of bug rather than individual instances of it.

During the work we identified security vulnerabilities in the platform and remediated them.

Outcome

Pages that took minutes returned in seconds. The platform stayed running throughout, and no rewrite was necessary.