SaaS
Platform architecture redesign enabling a Series B SaaS company to scale from 10K to 100K concurrent users with 99.95% uptime.
A Series B SaaS company — 120 engineers, $40M ARR, enterprise sales pipeline — was hitting architecture limits at 10,000 concurrent users. The platform had been built quickly across three years to chase product-market fit, and the architectural decisions that were appropriate at $1M ARR were now the constraints at $40M.
The specific problems: a relational database handling both transactional and analytical queries was showing 3–4 second response times under peak load; a deployment model where three product teams shared one release pipeline was serialising deployments and limiting release frequency to twice a week; and an authentication architecture that didn't support the SSO and role-based access control requirements of enterprise clients in the sales pipeline.
The stakes: two enterprise contracts representing $4M ARR were pending procurement sign-off. Both had raised architecture questions in due diligence that the current system couldn't confidently answer.
The engagement started with an Architecture Review to establish an accurate picture of the current system — not the intended design, but what had actually been built. The review produced a dependency map, a risk register with 14 findings ranked by severity and remediation priority, and a remediation roadmap sequenced to address the enterprise sales blockers first.
The Platform Engineering engagement that followed addressed the three highest-priority findings in sequence. The database was split: transactional queries remained on the primary, analytical and reporting queries were migrated to a read replica with a dedicated connection pool. The deployment pipeline was decomposed into three independent pipelines — one per product team — using feature flags to coordinate cross-team releases without serialisation. The authentication layer was redesigned to support SAML 2.0 SSO and a role-based access control model that satisfied the enterprise procurement requirements.
The platform reached 100,000 concurrent users in load testing before the enterprise contracts closed. Both enterprise clients completed technical due diligence successfully — the architecture questions that had stalled procurement were answered with documented evidence from the load tests and the architecture review deliverables.
Post-deployment: p95 API response time under peak load dropped from 3.8 seconds to 340ms. Release frequency increased from twice weekly to daily per team. The platform maintained 99.95% uptime across the following two quarters, including a product launch that drove a 4× spike in concurrent users over 48 hours.
The $4M ARR in enterprise contracts closed within six weeks of the architecture work completing.
Book a 30-minute strategy session. We'll walk you through relevant track record from your sector.