Client stories
Evidence from the readout room
These notes come from application owners and operations leads who commissioned stack health work on internal business apps — not anonymous star ratings.
The assessment caught a shared reporting extract that stalled three internal apps every month-end. We had treated each stall as a separate payroll or inventory issue for almost a year.
I wished the discovery interviews had been shorter — two of our owners were traveling — but the dependency map they produced was the first artifact our release managers actually pinned on the wall.
After the friction review we stopped stacking four deployments on the same Thursday night. Weekend pages dropped enough that our on-call rotation could breathe again.
Quarterly checkups keep the remediation sequence honest. Last quarter we slipped on the observability items; the re-score made that visible before another busy season.
They refused to invent a single health percentage for the board. Instead we got ranked hotspots and owners. That was harder to present, and far more useful in budget talks.
Extended note: month-end stall at a regional distributor
A Busan-based distributor brought us in after payroll, inventory, and leave-request systems each logged “slow month-end” tickets for nearly a year. Separate vendors had already inspected each application. Nothing durable changed.
During discovery we aligned ninety days of informal uptime notes on one timeline. A shared reporting extract — owned by nobody in the CMDB — restarted under load every month-end and starved connection pools used by all three apps. The Stack Health Assessment ranked that extract above a dozen application-local defects.
Remediation in the first thirty days was unglamorous: name an owner, add a health check, move the extract off the peak window. Later quarters tracked whether those three apps still clustered on the same nights. They did not. The mild reservation from their operations lead remains fair: discovery interviews ran long while two owners traveled. The map still became the artifact release managers kept.