Technical Rescue — Failed Agency Project Recovered in 3 Weeks

A UK property technology startup arrived with a situation we'd heard before, but rarely this severe. Nine months. Three agencies. £210,000 spent. Nothing in production. The most recent agency had gone quiet two weeks earlier after delivering a staging environment that crashed on login. The founder had a demo with a major UK estate agency group in 23 days — a meeting that had taken six months to arrange and represented the company's first real shot at a lighthouse client. Without a working product, there was nothing to demo. Without the demo, the next funding conversation became very difficult. The £200K+ already spent would be a sunk cost with nothing to show for it. We took a 72-hour assessment engagement before committing to anything. Our rule on rescue projects: don't promise a delivery date until you've read the code. What we found was a mixture of the abandonable and the salvageable — three distinct codebases from three different agencies, no shared conventions, no tests, three different database schemas that partially overlapped, and one module (the property matching algorithm, ironically the most business-critical piece) that was actually well-written and reusable.

• £210,000 spent across 9 months and three agencies with nothing deployable — the founder was 72 hours from cancelling the demo and potentially the company
• Three separate codebases with incompatible conventions, three partial database schemas, and no documentation — the codebase archaeology itself was a two-day project
• 23-day hard deadline for a demo with a flagship enterprise client that had taken 6 months to arrange
• One reusable asset in 210K lines of code: the property matching algorithm module from Agency 3 — everything else was either abandoned or actively broken
• The founder had lost all trust in technical vendors — any new commitment required extraordinary credibility, not just promises

We ran a 72-hour triage before making any commitments. The triage produced a written assessment: what was salvageable, what had to be rebuilt, and a realistic timeline with specific milestones the founder could hold us to. We committed to the demo deadline only after we'd read the code.

• 72-hour triage: full codebase audit, database schema analysis, and a written salvage/rebuild recommendation delivered before any development started
• Rescue sprint (Weeks 1–3): salvaged the property matching algorithm, rebuilt authentication, API layer, and core property search on a single clean PostgreSQL schema — demo-ready on day 19 of 23
• Demo environment hardened: synthetic data loaded for 50,000 properties across 12 UK cities, sub-500ms search responses, load-tested to 200 concurrent users
• Parallel documentation: every decision made during the rescue sprint was documented in real time — the founder could read exactly what we built and why at any point
• Full rebuild (Weeks 4–11): post-demo, we rebuilt the complete platform on the salvaged architecture with full test coverage, CI/CD, and production infrastructure
• Handover included a 2-hour recorded codebase walkthrough and a 30-day support period — designed so the next technical hire could onboard in under a week