Browser performance for advertising or tracking

Is there data on how much the average website today busies a web browser with things like advertising or tracking/analytics? For example, what percentage of RAM use or HTTP traffic is down to tracking/ads?

Short answer: The difference between low and high usage is so big that average or median makes no sense. You have to consider breaking the sites into segments and analyze segments. If you don't, you will see that marketing/analytics has virtually no impact on average. The global average/median impact is so low, that any stats on it would be either obvious or misleading.

But even the "high usage" of analytics and marketing is often barely noticeable or significant. This is mostly because of how powerful our hardware is and how little marketing/analytics need to track things and pass the data over. You would think that the data implies size, but it doesn't. Even here, on SO an analytics call is about a thousand times smaller than the size of a page, not counting all the core JS and supplementing JS libraries. In vast majority of cases, marketing and analytics are wrong places to go for page speed improvements even when lighthouse frowns upon "unused JS" in your favourite tag management system.

But I got curious and so let's dive a bit deeper?

Base

After a quick search, it appears that:
There are about 1 billion websites on the web.
There are about 28 million of websites that use the most primitive and easy analytics: google analytics.

If you use the 1 billion as a base, then the amount of RAM/network an average site uses is miniscule. Let's not do that.

Let's take the 28 million of GA websites plus about 10% over that as the base because GA is almost at 90% market share (as pretty much the only free solution that hosts your data). Besides GA, there are other analytics systems. There's the open-source Matomo which people don't use mostly due to how difficult it is to host your own analytics solution. There are things like Adobe Analytics, which is the next step in analytics solutions; it's used mostly by fortune 500 corps. And there are less significant systems like snowplow, tealium, ensighten and so on. Often GA is used in parallel with the others.

Basic Tracking

Vast majority of sites would just have the very basic tracking consisting of either analytics.js or gtag.js. They both are about 80 kb. Our library is loaded once per site and it is kept cached. Amount of memory they use is probably comparable to their size since they only do whatever is needed to send a very basic tracking info on every pageview in an async, non-blocking way. And they would typically send one network request per page or history change in case of gtag.js. And that non-blocking network request would be about 60 bytes. Not kilobytes. Bytes. Here's an example from SO:

SO's tracking is not very elaborate, but it's definitely more than default tracking and default is definitely median out there due to how much harder it is to shift from default to custom tracking.

Other analytics systems wouldn't be much different in these terms.

Basic tracking conclusion:
Out of a billion sites, 28 million use for marketing/analytics:

• About 100kb of memory usage
• About 80kb of network usage on the initial hit
• About 60 bytes of network usage on the subsequent pageviews

Low-effort tracking

Good, we're done with basics. Now let's get to our next stop: GTM. Or Junior-Mid level analytics. Again, Google's stats show that about 6.3 million sites use GTM. Most of them are likely have just a few tags published or are just staying empty, or were created to play around, or were created via the API during the configuration of a CMS and never touched again.

GTM's size can range since it's a very dynamic library and heavily depends on the number of entities used in it. Every time you add tags, triggers, variables, or change stuff, it requires you to "Publish" the changes, which actually implies rebuilding the library and the library's endpoint starts serving an updated version. The size varies from about 100 to 200kb. However, GTM is capable of dynamically loading other libraries on demand from third party endpoints.

GTM is a TMS (tag management system), and it has competitors. GTM's market share is similar to GA's given how GTM is quite comfortable and free to use. It's share is around 85% and we can safely presume that if a site has a TMS, but it's not a GTM, it's now an advanced set up.

How many GTMs can be considered to be loading extra scripts? Most people would just force GTM to load gtag.js or analytics.js, which we've went through in basics. Now, after gtag.js or analytics.js, the most popular analytics library to be loaded via GTM is most likely the facebook pixel library.

Looks like facebook is reluctant to share its pixel usage statistics, but there are estimations that land in around 2 million of sites. Sure, some sites would just use it from JS and ignore GTM, but it's hard to estimate how many do that and it's definitely not a good practice to delegate your marketing tracking to your front-end developers. Let's be conservative in our estimations and err on the side of caution. Let's presume all 2 million sites use GTM in parallel or serve their FB pixels through GTM.

Medium tracking conclusion:
Out of 28 million sites that use GA, 6.4 million add the following usage:

• About 200kb more of memory usage (due to GTM)
• About 100kb of network usage on the initial hit (due to GTM)
• About 60 bytes of network usage on the subsequent pageviews

Out of 6.4 million GTM sites, 2 million add the following usage:

• About 100 kb for fbevents.js and whatever it downloads. Only on the initial pageview.
• About 50 bytes per pageview
• At this point, GTMs would likely have additional tracking, so add 50bytes for every scroll, and meaningful CTA click.
• Add 100 bytes per e-commerce action: add/remove from cart, checkout steps, pdp view, purchase.

High-effort tracking.

This is basically large to enterprise businesses that easily spend over $100k per month to buy traffic. Hard to estimate the number of these sites without internal ad platforms stats.

Because they would buy traffic from various vendors at the same time, for multiple marketing campaigns running in parallel, they end up using more libraries and pixels. It's common to see about five different vendors tracking in parallel in addition to what we already have, so:

Out of 2 million of facebook pixel users, let's say for 1% of them add the following usage:

• About 500 kb for different third party marketing libraries, for the initial pageview only.
• About 250 bytes per pageview
• About 250 bytes for each meaningful user conversion. Mostly starts of the important funnels, then meaningful milestones in them and completions.

Now note that it may seem like loading a megabyte worth of libraries on the initial page load is likely to delay the page. It's not. Very unlikely. These libraries are loaded asyncly and usually the loading starts after the initial rendering of the DOM is done. Also, even a megabyte is nothing in terms of modern internet speeds. You can test it. Try loading sites with a bunch of pixels blocking the pixels' libraries and compare it to the normal loads. You'd be very unlikely to notice any significant difference.