Client stories
Specific notes from refactoring and tuning work — including where the engagement could have gone further.
“They refused to recommend a rewrite until we had a week of p99 traces. The first change was a connection-pool fix that cut checkout latency by almost half — less glamorous than a redesign, and exactly what we needed before Black Friday.”Priya N., engineering manager — performance tuning, Sydney retail platform
“The boundary workshop was useful, though I wish we had scheduled a second day for the payment domain. Still, the extraction plan for inventory services is the one our leads still follow six months later.”Marcus T., principal engineer — microservices refactoring, Melbourne logistics
“The architecture review named three risks our internal deck had softened. Board conversation was shorter because the document was specific about shared databases and release coupling.”Helen K., CTO — architecture review, Adelaide SaaS for trades
“Remote sessions worked. Diagrams arrived annotated the same day. The only friction was getting our database vendor on a call — that delay was ours, not WebRidge’s.”Jonah R., platform lead — performance tuning, Perth fintech
Extended note: inventory extraction, Melbourne
A logistics company had grown three “inventory” services that all wrote to the same stock table. Releases stalled whenever pricing and warehouse teams shipped in the same week. WebRidge ran a four-week refactoring engagement: discovery mapped the writers, workshops established a single stock-owner service, and the plan sequenced read-path cutovers before write-path moves.
Outcome after their team executed the first two stages: pricing releases no longer required warehouse approvals. The payment extraction was deferred — correctly, given staffing — and remains on their roadmap rather than being forced into the same quarter.
Extended note: quote API latency, Brisbane
An insurance aggregator’s quote API breached its p95 target during afternoon peaks. A two-week tuning consultation found synchronous calls to a rarely changing reference service and an N+1 pattern in eligibility checks. After their developers shipped caching and a batched eligibility query, afternoon p95 returned under the target for three consecutive weeks of monitoring.