What's the right way to prioritize which third-party scripts to keep versus cut when a performance audit flags too many running simultaneously?

Where do the third-party scripts on a Shopify store come from?

Third-party scripts reach a Shopify store by at least four routes, so an audit that checks only the admin's app list always undercounts. An inventory like this one answers how many scripts count as too many for your own store.

1. Theme code and app embeds. Hardcoded tags in theme files, including any a previous agency left, plus app embed blocks. The theme editor shows each app embed, so you can see an app running and turn it off (Shopify developer docs, October 2026).
2. Pixels in Customer Events. App pixels and custom pixels.
3. The tag manager. One entry in your inventory, often a dozen tags in the browser.
4. Storefront script tags. Apps have not been able to create or update them since 1 October 2026, and Shopify stops injecting them on 1 March 2027 (Shopify developer docs, October 2026).

Uninstalled apps are the fifth route and the one that pays fastest. Shopify says "Uninstalling an app doesn't automatically remove its code from your theme" (Shopify Help Center, October 2026), so old snippets keep loading with nobody owning them. What each route costs in load time is a separate question.

How do you score each script for keep or cut?

Each script gets scored on two axes, measured cost and a named owner, because cost says what to fix and ownership says what you may delete.

"Measured cost" means a number from a lab profile: main-thread time attributed to that script's origin, and whether it blocks rendering or first interaction. It is not a guess, and not the vendor's claim to be lightweight. "Named owner" means a person, not a team. If the answer to "who decides whether this stays" is a department, the script has no owner.

Who decides whether a script stays?

A script's named owner decides whether it stays, and the performance owner decides the budget it must fit. The rule we use is no owner, no script: a tag whose owner has left is cut by default, and anyone who objects must claim it before the audit starts.

Decision rights then split cleanly, so nobody overrides the budget by escalating. A budget that can be escalated past is a suggestion. Deploi's script governance record names the failure mode: thresholds get argued about, and "a budget nobody agreed to is a build failure everyone learns to override" (October 2026). So agree the number before the first script fails it, and write it where marketing can read it.

In what order should you cut scripts?

Cut scripts in this order: inventory, measure, assign owners, cut, move tracking tags, gate the budget and measure again, so owners are named before anyone sees the cost numbers.

1. Inventory every route: theme code and app embeds, Customer Events pixels, the tag manager and script tags, plus leftover code from uninstalled apps.
2. Measure the three templates that matter: home, the top collection and the top product page.
3. Assign owners before showing anyone the cost numbers.
4. Cut the scripts nobody owns in a preview theme, then publish.
5. Move surviving tracking tags to Customer Events where the vendor supports it, because app pixels run in web workers, background scripts with no access to the page (Shopify developer docs, October 2026).
6. Set the budget and gate it. Shopify's reports annotate app installs and theme updates over 90 days (Shopify Help Center, October 2026), which catches a regression only after the fact.
7. Measure again at week two, because app-injected scripts land without a deploy.

When an app swaps its script tag for an app embed, check that only one runs: Shopify warns that both together load the script twice (Shopify developer docs, October 2026).

Which scripts should you not cut?

Some scripts should not be cut on cost alone, because they carry a legal duty, a checkout dependency or a live campaign.

What does performance monitoring cost?

Performance monitoring costs $75 to $90 a month at entry tiers, with SpeedCurve's $90 on annual billing, so price is rarely the reason to skip it. Calibre's Starter plan is $75 a month with up to 5,000 real-user sessions and 5,000 synthetic tests, and Team is $150 (Calibre, October 2026). SpeedCurve prices plans by the real-user page views and synthetic checks you choose. On annual billing, which includes a 20% discount, Starter starts at $90 a month and Growth at $576, with Enterprise custom (SpeedCurve, October 2026). Billed monthly, those plans come to about $113 and $720 (Deploi arithmetic).

Both tools add a script to the page you are trying to keep light. Deploi's performance monitoring record calls that "The category's core irony" (October 2026), so budget the monitoring script inside the same limit as everything else.