Client stories
Evidence from rooms where events actually broke
These accounts name the kind of work we did and what remained unfinished. They are not star ratings or marketplace widgets.
During the Event Flow Assessment we discovered our “single order stream” was three republish hops with different IDs. The written map embarrassed a few of us — usefully. We still have not instrumented the middle hop the way they recommended, so late cancels remain a soft spot.
Sora P., platform engineer, marketplace company
The tuning sessions were blunt about alerts we loved that meant nothing. After the third meeting our overnight pages dropped. I would have liked one more session on weekend traffic patterns; we ran out of calendar before the Chuseok peak.
Daniel K., on-call coordinator
Desk review turnaround was faster than our internal architecture forum. Their memo caught a missing dead-letter path we had sketched but never owned. We implemented half of the notes before launch; the other half waits on a budget cycle.
Hyeon-woo L., solutions architect
Case note: inventory deltas across two warehouses
A Gyeonggi retailer asked us to examine why storefront stock disagreed with warehouse counts during flash sales. The Real-Time Event Flow Assessment traced deltas through a fan-out that dropped messages when a secondary warehouse republisher restarted.
Findings prioritized a consumer-side applied counter and a restart playbook. Six weeks later the client reported fewer customer complaints about “in stock” items that were not. They also noted that weekend staffing still struggles to interpret the new counters — a training gap outside our original scope.
Case note: quiet drops after a schema change
A fintech operations group booked an Incident Pattern Retrospective after three separate “missing notification” tickets. Pattern work showed each incident followed a schema field becoming required on one producer while an older consumer still filtered unknowns.
The action list assigned schema-change observation to the producer guild and a compatibility check to the consumer owners. Not every ticket closed in the same sprint — one consumer team disputed ownership for two weeks — but the shared language stopped the incidents from being treated as unrelated flukes.
Want a similar look at your paths?
Share a recent event incident or a release you are nervous about. We will suggest an engagement that matches the evidence you already have.
Start an inquiry