Migration & observability
Bounteous — Knitwell Group commerce platform
View as table
| Measure | Week 0 | Week 3 | Change |
|---|---|---|---|
| Tickets per week | 161 | 2 | −98% |
The problem
Five core services — OMS, PIM, Offers, Integration, and Payment — plus their microservices needed to move onto a new platform without a minute of downtime, on a commerce system running $2.5B in GMV. Inherited monitoring was noisy enough that real signals were getting buried.
The approach
Led hypercare across 35+ daily tickets while coordinating engineering, operations, and DevOps. Rebuilt the alerting architecture: 54 new health-check and anomaly monitors went in, and 250+ inherited monitors were rationalized down to 150. Authored the Datadog runbooks for the critical microservices — EKS, MongoDB, SQS, Identity, OMS — so the noise reduction would hold.
The impact
Ticket noise on critical services dropped from 161 to 2 tickets per week within three weeks. Systems stabilized inside the 7–10 week hypercare window, with zero downtime on the migration itself.