A fast homepage does not mean a fast store. Product variants, filters, third-party scripts and checkout can all feel different on a customer’s phone. This guide is for ecommerce and digital product teams choosing a web performance agency to improve actual shopping journeys, rather than buying another colourful audit deck.
What a credible proposal should include
- A baseline for mobile and desktop across homepage, collection, product and checkout journeys.
- A clear separation of real-user field metrics and lab diagnostics, with data gaps disclosed.
- A ranked backlog showing expected user impact, implementation risk, owner and validation method.
- Regression testing for checkout, consent, accessibility, search and measurement tags.
- A handover plan with performance budgets, monitoring and post-launch ownership.
Measure the buying journey, not just a score
Google’s Core Web Vitals cover LCP for loading, INP for responsiveness and CLS for visual stability. The recommended ’good’ thresholds are LCP of 2.5 seconds or less, INP of 200 milliseconds or less and CLS of 0.1 or less at the 75th percentile. These are experience benchmarks, not promised revenue or ranking outcomes.
Ask for segmentation by device, template and journey. A product configurator can be painfully slow even when a homepage passes a lab test. If a site lacks enough field data, the agency should explain the limitation and use lab work to diagnose likely bottlenecks without presenting estimates as observed customer behaviour.
The brief changes with your architecture
- App-heavy Shopify store: audit third-party scripts, theme choices and the business purpose of each app.
- Headless storefront: investigate rendering, API latency, caching, image delivery and slower network conditions.
- Large catalogue with dynamic filters: test interaction delays, product cards, loading states and layout movement.
Buy implementation capacity, not only analysis
An SEO specialist may spot a problem but need frontend or platform engineering to fix it. Separate discovery, code changes, QA, release and monitoring in the scope. If the agency cannot implement, ask who will and what the handoff contains. A PDF-only deliverable can be valuable, but call it an audit rather than a performance improvement engagement.
Third-party scripts require care. Analytics, personalisation and advertising tags should not be removed without involving the owners of consent, measurement and commercial workflows. A CRO agency may help evaluate conversion hypotheses, but site-speed remediation and experimentation are distinct projects.
Seven questions for the shortlist
- Which templates and user segments will you investigate first?
- How do you combine CrUX or other field data with controlled lab tests?
- Which fixes can your team ship directly?
- How will you isolate images, JavaScript, app overhead and server response?
- How do you test checkout, search, variants and consent after a release?
- Who owns ongoing measurement and performance budgets?
- Which optimisations would you defer, and why?
Pricing and acceptance criteria
Request separate estimates for diagnosis, implementation and verification. Identify dependencies on your platform vendor, engineering team and release schedule. Set measurable acceptance criteria per template, and agree how much field traffic is needed before evaluating changes. Avoid asking for a guaranteed uplift that nobody can substantiate.
Red flags and the next move
Be cautious about guaranteed search positions, one-size-fits-all speed scores and claims that a single plugin fixes everything. A proposal without a test environment, access requirements or named technical owners is incomplete. If you actually need a platform rebuild, start with an ecommerce development partner instead. Otherwise prepare a short brief with your storefront architecture, critical journeys and decision owners.
Technical references: Google Web Vitals and Google’s performance optimisation guidance.