Why an MVP is slow: diagnose before optimising

A slow MVP needs a measured diagnosis before a new hosting plan or framework.

A slow MVP needs a measured diagnosis before a new hosting plan or framework. Start with one user-visible symptom: a page takes too long to become usable, a search stalls or an import blocks other work. Record where, when and for whom it happens. That turns a vague complaint into a journey the team can reproduce.

Separate browser and server time

Inspect document loading, assets, rendering and interactive work in the browser. Correlate requests with server traces and logs using safe identifiers. A fast API cannot compensate for a huge client bundle, while a lightweight page can still wait on a slow database. Compare affected devices, networks and data sizes rather than testing only a developer laptop.

Locate queueing and repeated work

Measure database query duration, query count, connection waits and external calls. Look for sequential operations that could be bounded or combined, but preserve ordering where business rules require it. Check background jobs competing for the same resources. Tail latency and throughput under representative load reveal bottlenecks that a single request may not show.

Choose one hypothesis at a time

For example, if response time grows with result count, inspect pagination and query plans. If only first visits are slow, inspect cold starts and initial assets. If imports disrupt checkout, inspect shared resource contention. State the expected improvement before applying an index, cache or scheduling change. Keep a rollback path and avoid treating every plausible optimisation as necessary.

Verify user benefit and correctness

Repeat the same journey and workload after the change. Check error rates, data freshness, permissions and background completion as well as speed. Add a focused regression budget or alert for the repaired path. Document the remaining constraint so the next change begins with evidence. Optimisation is complete when the agreed experience improves without creating an unacceptable correctness or operating cost.

• Related service →
• Recover a failing software project: the first two weeks →
• Audit an AI-generated MVP before launch →

• Google SRE — Reference documentation →