Ecommerce site speed: measure the complete buying task

Measure ecommerce site speed through product choice and checkout, then test the image, script, data or cache change that addresses the observed delay.

Ecommerce site speed improves when you identify the work delaying a customer’s buying step, change that work and repeat the same task. Follow a product choice through to a verified checkout result. A fast initial page can still leave a customer waiting for the selected size or a basket update.

For a retail team, the starting point is a repeatable buying task and access to browser diagnostics. Make changes on a test copy of the store, with test customer records and the payment provider’s test mode. Keep a recoverable copy of the previous configuration.

Measure the actual buying task

Measure the delay between the customer’s action and the result they need before making another decision. For a size change, that result includes the selected option’s price and availability. For checkout, it includes a confirmed test order.

  1. Choose a representative product with variants. Write down the starting variant, the option to select and the expected product facts. Include an unavailable option as a separate case.
  2. Add the available variant to the basket, change its quantity and proceed to checkout. Use a repeat customer and a guest when both journeys matter to your store.
  3. Repeat on the phones and desktop devices your customers use. Record the browser version, connection conditions and whether the visit starts with an empty or populated cache.
  4. Capture the slow step in the browser’s Performance panel. Keep a separate Network recording for the request that retrieves the product or basket result. Record the visible symptom beside the request timing.

Chrome DevTools’ runtime performance guide explains how to record activity and investigate work on the browser’s main thread. That is the thread responsible for much of the page’s scripting and rendering. If the button appears unresponsive, inspect what occupies it after the click.

Lab measurements reproduce a chosen device and connection setup. Field measurements come from real visits and describe the conditions those visitors experience. Use field evidence to find affected routes or devices, then recreate a comparable slow case in the lab.

Google’s Core Web Vitals definitions separate loading, responsiveness and visual stability. Largest Contentful Paint (LCP) measures when the largest eligible image or text block appears. Interaction to Next Paint (INP) measures interaction responsiveness. Cumulative Layout Shift (CLS) measures unexpected layout movement.

Google’s good thresholds are LCP at 2.5 seconds or less, INP at 200 milliseconds or less and CLS at 0.1 or less. Assess them at the 75th percentile, separating mobile and desktop visits. These are page-experience measures. Add a task-specific measurement for a variant request or order confirmation that finishes after the next paint.

Identify the responsible work

Match the visible wait to the browser trace and the request that supplies the missing result. A large image and a slow stock lookup need different corrections.

Diagnose the wait by its evidence

Observed symptomEvidence to inspectFocused correction to investigate
The product photo arrives lateImage request start, transfer size and decoding workCorrect the requested image size or loading priority
A click waits before the page respondsLong scripting work around the interactionRemove or reduce the responsible script work
Content has arrived but paints lateStyle, layout and paint work in the traceReduce the specific rendering work delaying display
The selected size stays on a loading stateProduct request timing and returned variant dataInvestigate the store endpoint or its upstream dependency
A repeat visit is fast but shows an old priceCached response and the store’s update pathRepair invalidation while preserving current prices

Image optimisation for ecommerce starts with the image the customer is waiting for. Compare its delivered dimensions with its displayed size. Compress an oversized file and serve a suitable responsive version. Preserve enough detail to judge the product.

If the image request starts late, reducing file size alone leaves that initial wait untouched. Google’s LCP diagnostic guide separates request delay, download time and the delay before rendering. It also explains why the LCP image must not be lazy-loaded. Apply lazy loading to suitable off-screen media instead.

Use the Network panel’s timing tools to distinguish waiting for a response from downloading it. A completed response with a frozen selector points back towards browser work. A pending stock request calls for investigation of the request and server-side processing. A hosting change needs evidence that hosting resources cause the wait.

Apply the diagnosis to WooCommerce

WooCommerce performance optimisation includes the theme and extensions that participate in the buying step. On an isolated store copy, compare the failing task with the suspected extension disabled. If needed, compare with a supported default theme. Preserve the same product data and test conditions so the comparison answers a useful question.

WooCommerce’s performance guidance covers responsive images and code optimisation alongside caching. Use those as areas to investigate, not a reason to install several optimisation plugins at once.

Caching requires a separate correctness check. WooCommerce’s cache configuration guide says Cart, Checkout and My Account must remain dynamic because they contain customer-specific information. A cached basket that belongs to another session is a failure even if it loads quickly.

Test the focused correction

Change the smallest part that explains the observed delay, then repeat the original buying task under the recorded conditions. Keep the previous version available so you can compare or recover it.

Consider this hypothetical diagnostic case. A shoe’s size selector updates its label immediately, but its price stays unchanged while a variant request waits. The Network recording shows that request waiting for the server response. Compressing the product image does not address the waiting request. Investigate the variant endpoint and the extension work it invokes before changing image delivery.

After correcting the responsible work, run the same size selection again. Compare the time until the correct variant facts appear. Repeat enough runs to see whether the change survives normal variation, retaining all results rather than selecting the fastest one.

Then test the nearest failure case in the same isolated store:

  • Select an unavailable size and confirm the store prevents its purchase.
  • Change the selected variant and confirm the basket contains that exact option.
  • Update a test product’s price and stock, then check both a fresh visit and a repeat visit receive the new facts.
  • Complete checkout with test payment details and confirm the order has the expected items and total.
  • Open a second customer session and confirm its basket and account data stay separate.

A failed request must leave a recoverable state. Interrupt the variant lookup and check that the customer sees a usable retry or error. The store must not silently keep the previous size’s price under the newly selected label.

Keep the result interpretable

Report the corrected buying step with its test conditions and correctness results. Keep lab findings separate from field measurements, and identify whether the change affects the first visit or a cached repeat visit.

A useful record names the route and test product, device, browser version, connection profile and starting basket state. Include the before-and-after timings from comparable runs and the evidence tying the delay to the changed work. State which payment mode and checkout outcome you verified.

Recheck the affected task after a theme update, a new extension, a changed product-data connection or a cache-policy change. Each can invalidate the conditions behind the earlier result.

A shorter delay establishes a technical improvement in that setup. A sales effect needs a separate conversion rate optimisation assessment with comparable traffic and a defined outcome. Start by confirming that customers can choose the right variant and complete the purchase with current facts.

Give your store its own shopping agent

Customers ask in their own words. Your agent finds, compares and checks out from your catalogue, right on your site.

See it in actionFree trial · Simple setup