Let’s talk

Web & mobile · Site speed

Online store speed: what actually affects sales and what to fix first

A perfect PageSpeed score is not the goal; shoppers who reach the order confirmation are. In most online stores the speed that matters comes down to four things — images, scripts, third-party tags and the checkout — and they are not equally worth your time.

Measure what shoppers feel

There are two kinds of speed data, and they answer different questions.

Lab data is one simulated page load on a simulated phone: Lighthouse, or the score at the top of PageSpeed Insights. It is good for debugging and for checking a fix before release.

Field data is what your real visitors experienced, collected by Chrome and shown in PageSpeed Insights and Google Search Console. This is the number to manage. Google’s Core Web Vitals call a page “good” when, at the 75th percentile of visits, Largest Contentful Paint (LCP) is 2.5 seconds or less, Interaction to Next Paint (INP) is 200 milliseconds or less, and Cumulative Layout Shift is 0.1 or less.

Measure by template, not by the home page. In most stores, product and category pages carry the traffic and the cart and checkout carry the money; the home page is often the fastest and least important of the lot. Split by device too: desktop is usually fine, and mobile is where the problems are.

Then do it by hand. Take a mid-range Android phone on cellular, not office Wi-Fi, and time three moments: from tapping a category to seeing the first product photo, from tapping “Add to cart” to the cart updating, from tapping “Pay” to the confirmation. Those three waits are what shoppers remember.

Images: the heaviest part of the page

On a product page, the main photo is usually the LCP element, so it decides the most visible speed metric. Typical problems:

  • studio photos uploaded at 4,000 pixels wide and shrunk by the browser to 400;
  • one file for every screen instead of responsive sizes;
  • JPEG and PNG where WebP or AVIF would be a fraction of the weight;
  • the main product image lazy-loaded, so the browser fetches it last;
  • no width and height set, so the layout jumps when the image arrives.

Fix the main image first: right size, modern format, loaded eagerly and with high priority. Lazy-load everything below the first screen. On a category page with, say, 48 products, the thumbnails should be sized for the grid, not for the zoom view. If the home page has a carousel, the first slide is the LCP element, and the other four slides should not load before it.

Scripts and third-party tags

This is where most stores lose the most time, and where nobody feels responsible.

Every app and tag adds JavaScript: reviews, upsells, wishlists, chat, loyalty, pop-ups, heatmaps, three ad pixels, an A/B testing tool. All of it runs on the phone’s processor, and a mid-range phone is far slower at this than a laptop. Heavy scripts show up as poor INP: the shopper taps “Add to cart” and nothing seems to happen.

Do an inventory. List every tag in the tag manager and every script in the theme. For each, write down the owner, the purpose and the last time anyone used its data. Uninstalled apps often leave their code behind in the theme.

Remove what has no owner. Then delay what isn’t needed for the first interaction: load chat when the shopper opens it, run heatmaps on a sample of visits, fire marketing pixels after the page is usable.

Watch for hidden pages. Some A/B testing tools hide the whole page until their script loads, to avoid flicker. That delays every visitor, including the ones who are not in a test.

Check filters and search. A category filter that reloads the full page on every checkbox feels slow even when the score looks fine.

Checkout: speed you can’t see in a score

Checkout speed is about steps and waits, not page scores. Each step often calls another system: shipping rates from carriers, address validation, tax calculation, stock checks against the warehouse, the payment widget. Each call is a pause the shopper sits through, and none of them can be cached the way a product page can.

  • Offer express wallets (Apple Pay, Google Pay, Shop Pay or your platform’s equivalent) so returning shoppers skip typing.
  • Keep guest checkout; asking for an account before payment adds a step.
  • Strip out marketing scripts the checkout doesn’t need.
  • Track the time between steps and the error rate at the payment step, not just the conversion rate.

Place a real order on a phone once a week. It is the cheapest monitoring you will ever set up.

What to fix first

A practical order, from most to least money per hour of work:

  1. Anything broken or blocking in checkout: errors, timeouts, a payment widget that loads slowly.
  2. The main image on product and category pages.
  3. Third-party tags nobody owns.
  4. Interaction delays on “Add to cart,” filters and search.
  5. Everything else, including the last few points of the lab score.

Ship one change at a time and annotate the release date in analytics. Otherwise you will never know which fix moved conversion and which only moved the score.

What to do this week

  1. Open PageSpeed Insights for a top product page and a top category page, and write down the mobile field LCP and INP.
  2. Export the list of tags and apps and mark each with an owner and a purpose.
  3. Check the file size of the main image on your best-selling product page.
  4. Place an order on a mid-range phone over cellular and time each step.
  5. Pick one fix from the list above, ship it and annotate it.

Boson Group builds online stores and runs the performance marketing that brings traffic to them, so we look at speed together with what each visitor costs. If you want a speed review of your store with a ranked list of fixes, write to team@boson.group.

More about Web & mobile at Boson Group

Contact

Send a brief — we reply with a time and cost estimate within 24 hours.

Brief

Tell us about the project.

Send a brief — we reply with a time and cost estimate within 24 hours.

What do you need?

* required