← Back to the blog

Performance

How to find which apps are slowing down your Shopify store

Don't guess from the app list. Measure what each app costs in the browser, turn them off one at a time on a duplicate theme, and keep only the ones that pay for their JavaScript.

The fastest way to find which apps are slowing down a Shopify store is to stop guessing from the app list and measure. List the third-party scripts that load on your key templates, then turn each app off on a duplicate theme and measure again. The app whose removal moves the number is your answer.

Wellness and supplement stores feel this more than most. The usual stack is reviews, subscriptions, loyalty, a product quiz, bundles, upsells, chat, a pop-up and a few analytics tools, and each one wants a script on the page. The theme modernization I wrote up here started with a theme carrying more than 40 apps, installed and half uninstalled. Nobody had a clear picture of which ones still ran.

Where do apps load code in a Shopify theme?

Before you measure, know where to look. An app can reach your storefront in five ways:

  • App embeds. Turned on in the theme editor under App embeds. They load on every page, whether or not the page uses the app.
  • App blocks. Placed inside a section, so they only load where you put them. This is the cheapest way for an app to exist.
  • Script tags. Older apps register scripts through Shopify’s ScriptTag API. Shopify injects them through content_for_header, so you won’t find them in your theme code at all.
  • Code in your theme files. Snippets, render tags and <script> tags an app or an installer added to theme.liquid or a template.
  • Web pixels. Tracking set up under Customer events. These run in a sandbox, separate from the page, so they’re usually the least of your problems.

Step 1: start from field data

Open PageSpeed Insights for your home page, a collection and your best-selling product, and read the field data at the top, on mobile. Or use the Core Web Vitals report in Search Console if the site has enough traffic. You want to know which template fails and on which metric: LCP, INP or CLS.

This matters because the lab score moves every time you run it. Field data tells you whether real customers have a problem. The lab is where you debug it.

Step 2: list the third-party scripts on one template

Open the product page in Chrome, open the console and paste this. It groups every script the page loaded by where it came from.

const rows = {};
for (const e of performance.getEntriesByType('resource')) {
  if (e.initiatorType !== 'script') continue;
  const url = new URL(e.name);
  if (url.hostname === location.hostname) continue;
  // Theme app extensions are served from cdn.shopify.com/extensions/...
  const key = url.pathname.startsWith('/extensions/')
    ? `app extension: ${url.pathname.split('/').pop()}`
    : url.hostname;
  rows[key] ??= { scripts: 0, kb: 0 };
  rows[key].scripts += 1;
  rows[key].kb += e.transferSize / 1024;
}
for (const r of Object.values(rows)) r.kb = Math.round(r.kb);
console.table(rows);

Two caveats. A cross-origin server that doesn’t send Timing-Allow-Origin reports a size of 0, and so does anything served from cache, so do a hard reload first and treat the sizes as a floor. And bytes aren’t the whole story: a small script can still block the main thread. For that, record a load in the Performance panel of Chrome DevTools, which breaks main-thread time down by third party. That’s the number that hurts INP.

You now have a list of suspects with names on them.

Step 3: turn apps off one at a time on a duplicate theme

Never test on the live theme. Duplicate it, open the duplicate in the theme editor, and use its preview link for every measurement.

  1. Run Lighthouse on mobile three times against the preview and write down the median. That’s your baseline.
  2. Turn off one app embed, or remove one app block. Save.
  3. Measure again the same way, three runs, median.
  4. Turn it back on and move to the next app.

Script tags and code that lives outside the theme editor can’t be toggled there. For those, use request blocking in Chrome DevTools: right-click the script in the Network panel, choose to block its domain, reload and measure. It’s a fast way to see what the page would be like without that app before you touch anything.

Go one app at a time. Turning off five at once tells you the five together are slow, which you already knew.

Step 4: decide what to do with each app

Once you know what each app costs, you have five options, roughly from least to most effort:

  • Keep it. It’s expensive but it earns its cost. Write that down so nobody has to measure it again next month.
  • Scope it. If the app offers an app block, use the block on the template that needs it instead of an embed on every page. A reviews widget has no business loading on the cart page.
  • Delay it. Some widgets, chat is the classic one, can load on the first interaction instead of on page load. Check with the app that it supports this and that nothing it tracks breaks.
  • Replace it with theme code. An announcement bar, a size chart or a simple product tab doesn’t need an app. A few lines of Liquid and CSS cost nothing at runtime.
  • Remove it. Then do step 5, because uninstalling isn’t always the end.

Step 5: clean up what uninstalled apps left behind

Embeds and blocks from theme app extensions disappear with the app. Code that was written into your theme files doesn’t. In the code editor of your duplicate theme, search for the app’s name and its script domain, check theme.liquid and your snippets for render tags that point at things you no longer use, and delete what’s dead. Then measure one last time.

The rule I follow

Every app has to pay for its JavaScript. Measure what it costs, decide on purpose, and repeat the check every time someone installs something new.

Running this audit every quarter, and keeping new apps from undoing it, is part of what I do for wellness and supplement brands on a monthly retainer through MTB Labs.

FAQ

Do uninstalled apps still slow down a Shopify store?

They can. App embeds and app blocks from theme app extensions go away when the app is uninstalled, but code an app or an installer wrote into your theme files (snippets, render tags, script tags in theme.liquid) stays until someone deletes it.

Should I trust the PageSpeed Insights score?

Use the field data at the top of the report, or the Core Web Vitals report in Search Console, to decide whether you have a problem. Use the lab score to compare before and after while you test, always on mobile, and take the median of a few runs.

Which metric do apps hurt the most?

Usually Interaction to Next Paint (INP), because app scripts compete with the page for the main thread. Apps that inject banners or widgets above the fold can also push back LCP or cause layout shift.

Should I just remove every app I can?

No. Some apps make more money than their JavaScript costs. The goal is to know what each one costs, then keep, scope, delay, replace or remove it on purpose.