Diagnose it properly before changing anything.
Most performance work fails because it starts with a guess. Someone is sure it is the database, so the database gets optimised, and the page is still slow because the actual cost was three sequential requests that could have been one. Measure first, always.
Why people get in touch
Nobody knows why it is slow
Everyone has a theory and no instrumentation. The theories are usually wrong, and acting on them costs weeks.
The migration cannot stop the roadmap
A big-bang cutover means freezing feature work, which nobody will approve. So the migration never starts.
It has broken before
A previous attempt lost data, broke links, or took the site down. The organisation is now — reasonably — afraid of the change.
What the work includes
Instrumentation first
Real measurements before any change, so improvement is demonstrable rather than claimed.
Root-cause diagnosis
The actual bottleneck, which is often not where the team expects.
Incremental migration
Old and new running together, moving one piece at a time, reversible at every step.
Link and data preservation
Nothing that already works stops working — URLs, search rankings, integrations.
Regression tests
So the gain does not quietly disappear over the next two quarters.
Typical stack
Measurement
- Core Web Vitals
- Query profiling
- Flame graphs
- Real user monitoring
Frontend
- Bundle analysis
- Rendering strategy
- Caching
- Image pipelines
Backend
- Query optimisation
- Indexing
- Caching layers
- Job queues
Migration
- Dual-write
- Backfills
- Redirect mapping
- Feature flags
How it goes
A call, not a pitch
Thirty minutes on what you are actually trying to ship. If I am not the right person, I will say so and usually know who is.
A written scope
Before any money moves, you get the outcome, the approach, the price, and what is explicitly out of scope. No hourly surprises.
Work in the open
Branches you can read, demos over status reports, and a written update every day. You always know where things stand.
A clean handover
Documentation, tests, and a walkthrough with your team. The goal is that you do not need me afterwards.
Questions
Not sure if this is what you need?
Thirty minutes, no pitch. If it turns out you need something else — or someone else — I’ll say so.
Book a call