Almost every slow website we are asked to look at was fast on the day it launched. Nobody broke it. It accumulated. A chat widget for a campaign that ended, a heatmap trial from two years ago, a second analytics tool somebody wanted for one report, three advertising pixels for platforms you no longer run ads on. None of them felt like a decision. Together they are the reason your homepage takes four seconds to become useful.
The uncomfortable part is that this weight is invisible in the place people look for it. Your pages have not changed. Your hosting has not changed. The code your own developers wrote has not changed. What changed is everything marketing added through the tag manager, which is precisely the layer nobody audits.
Third-party scripts cost you in three ways, and only the first is obvious.
Connection cost. Every new domain your page contacts needs a DNS lookup, a TCP connection and a TLS handshake before a single byte of useful content arrives. That is roughly 100–300 milliseconds each, before the script has done anything.
Main-thread cost. Scripts execute on the same thread that responds to taps and clicks. This is why a page can look finished but ignore you for a second when you try to press a button — the browser is busy running someone else's code.
Cascade cost. Tags load tags. A single container can pull in a dozen further requests you never approved, and they change without telling you.
As a rough benchmark: a small business site running fewer than ten third-party scripts is usually fine. Past fifteen to twenty, they are almost certainly the largest single cause of your load time, and a neglected tag manager is typically the biggest offender on the whole site.
This is a meeting, not a technical project. Export the list of every active tag in your tag manager, add anything hard-coded into the theme, and put the list on one screen with the people who asked for them.
In our experience a first audit removes somewhere between a third and half of everything on the list, and nobody notices anything missing except the load time.
The tags that earn their place still do not need to load immediately.
Defer everything non-essential. Nothing that measures a visit needs to block the visit. Analytics, heatmaps and advertising pixels can wait until the page has rendered.
Use a facade for heavy widgets. Chat widgets, video embeds and map embeds are the usual culprits. Show a lightweight placeholder that looks identical, and load the real thing only when someone interacts with it. A chat widget that loads on click instead of on arrival often saves more than every other optimisation combined.
Respect consent properly. Tags that only fire after consent are both a legal requirement in many markets and a performance improvement for everyone who declines.
Nobody ever decided to make the site slow. They just each added one small thing, and no one was responsible for the total.
Measure the result with a tool that reports third-party impact specifically rather than a single overall score, and check real visitors rather than a lab test on a fast connection. The number that matters is how long it takes before a real person on a mid-range phone can actually use the page.
Put a tag audit in the calendar once a year and give one person the authority to remove things. It costs an afternoon, requires no redesign, and routinely takes a second or more off load times — which is more than most expensive performance projects deliver. Speed is not only something you build. It is something you defend.
We audit what your website actually loads, cut what nobody owns, and rebuild the rest to be fast.
Talk to us