Lightweight analytics
The weight of an analytics script rarely decides anything on its own, but it is one of the few things you can measure instead of arguing about. This page gives our numbers and explains why, for Core Web Vitals, whether the script touches the page layout at all matters more than kilobytes.
Free plan with no time limit. No card. Setup takes a minute.
Our numbers
Scroll the table sideways for the remaining columns.
| What | How much |
|---|---|
| Weight, brotli compressed | 1.6 kB |
| Weight before compression | 4.0 kB |
| Requests per pageview | One |
| External dependencies | None |
| How it loads | The defer attribute, no render blocking |
| Asking the browser about its properties | None |
| Forced layout recalculations | None |
The weight is measured on the file we actually serve, not on the source before the build. We give both numbers, because "compressed" without saying which compression means nothing: brotli gives a different result than gzip.
Why kilobytes are not the whole story
Downloading twenty extra kilobytes costs a fraction of a second and on a decent connection it disappears into the noise. An analytics script can nonetheless do harm in ways its weight does not show, and that is the real reason to look at it when you care about Core Web Vitals.
- every query about a size, a position or a computed style forces a layout recalculation, and a recalculation at the wrong moment moves content under the reader finger
- work done on the main thread delays the response to the first click, which is exactly what INP measures
- a script loaded from someone else domain adds a name lookup and a TLS handshake before the download even starts
- a script that fetches further files turns one request into a chain of requests
What this means for your site
If your page is heavy for other reasons, swapping the analytics tool will not save it and we are not going to promise that. The difference is noticeable where the rest of the page is already optimised, and on phones with a slower processor and a worse connection, where every piece of main thread work costs many times what it costs on a desktop.
While we are here: you can measure your own page speed with us without adding another tool. The script reports five metrics, among them largest contentful paint, interaction to next paint and cumulative layout shift, and the report shows percentiles rather than averages. A handful of very slow loads drags an average until it stops describing anyone.
What this page does not compare
There are no competitor weights here, because we did not measure them on our own hardware, and copying numbers out of someone else marketing material would be precisely what this page is meant to avoid. Where we compare ourselves to a specific tool, we do it on a separate page and cite the source.
Questions that come up most often
- Does the script block rendering?
- No. It loads with the defer attribute, so the browser fetches it in parallel and runs it only after the document is built. It holds up neither the first paint of content nor anything before it.
- Can I host the script myself?
- We do not offer it for self hosting. The script has to send events to our address, so a local copy of the file would save one name lookup and nothing else, while drifting out of step with the version we serve.
- Will content blockers block your script?
- Some will, and we do not try to get around it. The script is an ordinary file from our domain; it does not hide behind a proxy on the customer domain and does not pretend to be a first party asset. A blocked script means no event, and we do not fill such gaps with modelling.
- How big is the extra noscript tag?
- Forty three bytes, because it is a transparent one pixel GIF. It is only ever requested by visitors with JavaScript disabled, since that is the only case where a browser executes the contents of a noscript tag.
Measure it on your own site
Paste one line and check Core Web Vitals before and after. The free plan covers one site and shows the speed metrics at no extra cost.
Sources
- Core Web Vitals: metric definitions and thresholds, web.dev, 2026
Updated September 15, 2026