Start a speed review with the pages customers use and the problems they experience. A single homepage score cannot tell you whether the collection filter responds, the product gallery jumps or the cart remains usable on a phone.
Keep real-user data and lab tests separate
Shopify's web performance reports provide real-user measures of loading, responsiveness and visual stability. The summary uses the past 30 days; reports can be delayed by up to 36 hours and retain the last 90 days. The default reports use the 75th percentile. Filter by device, page type and URL to find the journey needing attention. Shopify web performance reports.
Largest Contentful Paint (LCP) describes loading; Interaction to Next Paint (INP) describes responsiveness; Cumulative Layout Shift (CLS) describes unexpected movement. Shopify labels LCP at or below 2,500 milliseconds, INP at or below 200 milliseconds and CLS at or below 0.1 as good. These are experience thresholds, not revenue forecasts.
A Lighthouse or PageSpeed lab run is a controlled diagnostic sample. Record its date, device profile and network assumptions, repeat under comparable conditions and investigate the actual issues. Lab Total Blocking Time does not establish field INP. Missing field data does not prove a pass or a failure. How Shopify measures performance.
Review a representative mobile journey
- Open a key collection, use its filters and reach a product.
- Load the product with its real fonts, imagery and app content present.
- Check the gallery, variants, navigation and cart behaviour in the authorised test environment.
- Watch for movement when fonts, banners, reviews or sticky controls appear.
- Use keyboard controls and reduced-motion settings where relevant.
Record what happened, not just the score. A control that is fast but difficult to use still needs attention.
Find the cause before removing functionality
Images: check the displayed size against the delivered file, responsive variants and whether the main image is delayed. Keep genuine product detail while avoiding a file sized for a much larger display.
Fonts and layout: check the fonts customers really receive and reserve space for media and dynamic sections. A fallback-font screenshot is not enough to rule out movement.
Scripts and animation: inspect long work, repeated rendering, third-party tags and the chat widget. Test the interaction they affect instead of assuming every animation is a problem.
Apps: identify which app contributes to the observed issue. Some apps work only in the backend. A missing storefront widget does not prove an app is unused, and uninstalling it may affect fulfilment or subscriptions. Use the billing and dependency checklist before proposing removal.
Ship a small change and compare carefully
Save a baseline, identify one cause, make the smallest useful change in a working theme and repeat the relevant journey. Keep original assets and a rollback path. The store owner approves a live theme release and any app changes.
After release, check the current customer surface, comparable lab samples and real-user reports as data accumulates. One improved score does not prove a stable improvement. A faster load does not establish more orders, better rankings or extra revenue without separate measurement.
Maintain the experience as the store changes
Review performance when major media, theme sections, apps or campaigns change. Keep a short log of the page, problem, owner, change, verification and follow-up date. Prioritise customer impact over a score target.
Our speed optimisation service starts with measured diagnosis. Shopify support and Snowball can include ongoing improvements within an agreed scope. The Czarina story documents practical store work, with commercial results kept separate from delivery.
Sources and review
Guidance checked on 7 October 2026. Features and commercial terms can change; confirm availability for your own account and scope. These planning checks are recommendations and do not replace an assessment of the actual implementation.