Research · November 2026
What the GetCookies banner costs a page: sizes and load timing
The GetCookies loader is 3.9 KB gzipped and runs synchronously, first in the page head, so Google Consent Mode defaults are set before any Google tag and known trackers can be held back. The banner itself (33 KB gzipped) loads asynchronously after the HTML is parsed. In our test on a mid-range phone over fast 4G, that delayed Largest Contentful Paint by 180 ms, added no Total Blocking Time and caused no layout shift.
By the GetCookies team at Getia AS, Oslo. Measured 2026-11-20 with widget 2.6.0. Last updated 2026-11-22.
Key findings
- The loader is 10.2 KB minified, 3.9 KB gzipped and 3.5 KB with brotli. The banner bundle is 33 KB gzipped (28 KB brotli).
- The loader is render-blocking by design: on a first visit it held up parsing for a median 180 ms on fast 4G mobile, 592 ms on slow 4G (PageSpeed Insights mobile throttling) and 53 ms on desktop cable.
- On the test page, Largest Contentful Paint moved by the same amount: +180 ms (fast 4G), +596 ms (slow 4G), +52 ms (desktop).
- Total Blocking Time did not change in any profile, and Cumulative Layout Shift stayed at 0: the banner floats over the page instead of pushing content down.
- The banner was visible after a median 833 ms on fast 4G, 2597 ms on slow 4G and 251 ms on desktop, after 4 widget requests (loader, banner, visitor region, banner configuration).
File sizes
Sizes of the production build of widget 2.6.0, measured with gzip level 9 and brotli quality 11 (1 KB = 1,024 bytes). The transfer column is what the browser downloaded in the timing test, including HTTP compression overhead. The files served from app.getcookies.co match this build.
| File | Minified | Gzip | Brotli | Transferred in the test |
|---|---|---|---|---|
| Loader (loader.js, synchronous) | 10.2 KB | 3.9 KB | 3.5 KB | 4.1 KB |
| Banner (widget.min.js, asynchronous) | – | 33 KB | 28 KB | 32.7 KB |
| Banner configuration (JSON) | – | – | – | 0.8 KB |
Load timing on mobile and desktop
Median of 15 paired first-visit loads per profile (p75 below each value). Each run loaded the same test page with and without the widget, so every "+" value is the difference the widget made.
| Profile | Parser blocked by the loader | Banner visible | Largest Contentful Paint | Total Blocking Time | Cumulative Layout Shift |
|---|---|---|---|---|---|
| Fast 4G, mid-range mobile | 180 msp75 180 ms | 833 msp75 846 ms | +180 msp75 +182 ms | +0 msp75 +0 ms | 0.00p75 0.00 |
| Slow 4G, mid-range mobile (PageSpeed Insights mobile) | 592 msp75 598 ms | 2597 msp75 2612 ms | +596 msp75 +598 ms | +0 msp75 +0 ms | 0.00p75 0.00 |
| Desktop on cable | 53 msp75 53 ms | 251 msp75 253 ms | +52 msp75 +56 ms | +0 msp75 +0 ms | 0.00p75 0.00 |
- Parser blocked by the loader
- How long the HTML parser waited for the loader to download and run. This is the cost of running it synchronously.
- Banner visible
- Time from navigation start to the first frame in which the banner is shown (its 300 ms fade-in starts then).
- Largest Contentful Paint
- Largest Contentful Paint with the widget minus the same page without it. On this text-only test page the delay equals the parser block; on real pages it depends on what else blocks rendering.
- Total Blocking Time
- Added Total Blocking Time: main-thread tasks over 50 ms caused by the widget. The loader runs in about 4 ms.
- Cumulative Layout Shift
- Cumulative Layout Shift of the page with the banner shown. 0 means nothing on the page moved.
Why the loader is synchronous, and what that costs
Google tags only read a Consent Mode default that is already queued when they start. A loader marked async or defer usually runs after Google Tag Manager or gtag.js, so their tags fire with no consent state. The same race applies to holding back Meta, TikTok and other trackers. That is why the install snippet has no async attribute and belongs first in the head.
The price is the download time of a 3.9 KB file on a first visit. On a fast connection that is tens of milliseconds; on a slow mobile connection it is the largest cost the widget adds. Everything after it — the 33 KB banner, the region lookup and the configuration — loads without blocking the page.
If you cannot accept a synchronous script, the GetCookies Google Tag Manager template sets the defaults in the Consent Initialization trigger instead; see the GTM guide and the three install options.
Method
- Build: the production bundles of widget 2.6.0, served gzipped from a local test server (
e2e/widget/server.mjsin the GetCookies repository). - Install:
<script src="https://app.getcookies.co/api/v1/widget/loader.js?id=DOMAIN_ID"></script>as the first script in the head, without async or defer, exactly as the dashboard prints it. - Browser: headless Chromium 151.0.7922.34 on an Apple M1 Pro, with network and CPU throttling applied through the Chrome DevTools Protocol for each profile below.
- Runs: 15 paired first-visit loads per profile (empty cache): the test page with the widget and the identical page without it. Banner configuration and visitor region came from fixed test responses.
- Metrics: navigation and paint timings, Long Tasks for Total Blocking Time, and layout-shift entries for CLS, read in the page; medians and 75th percentiles reported.
- Reproduce:
cd backend/app/static/widget && npm run perf:update. The script writes the JSON file offered for download below.
Throttling profiles
- Fast 4G, mid-range mobile: Chrome DevTools "Fast 4G" preset (9 Mbps down, 1.5 Mbps up, x0.9; 60 ms x 2.75 latency), 4x CPU slowdown
- Slow 4G, mid-range mobile (PageSpeed Insights mobile): Lighthouse mobile DevTools throttling (150 ms x 3.75 latency, 1474.56 Kbps down, 675 Kbps up, 1 Kbps = 1024 bit/s), 4x CPU slowdown
- Desktop on cable: Lighthouse desktop throttling (40 ms latency, 10240 Kbps down and up), no CPU slowdown
Limitations
- One small, text-only test page. On a real page the loader competes with your own scripts, fonts and images, so the LCP difference can be smaller (other resources block too) or larger (a slower first byte from our server).
- First visits with an empty cache only. On repeat visits the loader is usually served from the browser cache for up to 5 minutes, then revalidated.
- Emulated throttling on one machine, not field data from real visitors. Lab numbers are useful for comparing versions, not as a promise for every device.
- Banner configuration and region were served from a local fixture, so the time our API takes in production is not included.
Data
The raw measurements (every metric, median and p75, per profile) and the file sizes, exactly as the build scripts wrote them on 2026-11-20.
Cite as: GetCookies (Getia AS), "What the GetCookies banner costs a page: sizes and load timing", measured 2026-11-20, getcookies.co/research/widget-performance-2026.
Questions about these numbers
How big is the GetCookies script?
The loader you put in the page head is 3.9 KB gzipped (10.2 KB minified). It loads the banner, 33 KB gzipped, asynchronously. About 38 KB is downloaded before the banner shows, including the banner configuration.
Does GetCookies slow down my site?
It adds the loader's download time to the first visit, because it runs synchronously: +180 ms Largest Contentful Paint on fast 4G mobile and +596 ms on slow 4G in our test page. It added no Total Blocking Time and no layout shift.
Can I load the GetCookies loader with async or defer?
It still works, but it usually runs after Google Tag Manager or gtag.js, so Google tags start without a consent default and trackers may run before they can be held back. The banner then logs a placement warning. Use the GTM template if you want the defaults set from Tag Manager instead.
Will the banner hurt my Core Web Vitals?
The banner itself caused no layout shift (CLS 0) and no long tasks in the test. Largest Contentful Paint moves by the loader's download time on a first visit; on real pages that depends on what else blocks rendering.
Last updated 2026-11-22. All research
Try GetCookies on your own site
Start on the Free plan: one domain, a cookie scan and a consent banner, no credit card required. Upgrade only when you need more.
Start free