The good old days
When Google Analytics first came along, back in 2005, it was genuinely revolutionary. Before that, getting a handle on your website traffic was either expensive (having to pay for third party tools) or involved squinting at your server logs.
It genuinely gave any business, small or large, a window into who was visiting your site, where from, and what they did when they got there. It democratised web analytics overnight. It was truly groundbreaking.
But as time went on...
That position was slowly eroded. Ad blockers were created to prevent tracking. Web browsers (not Chrome, obviously) started to build-in privacy features. GDPR and cookie consent requirements reduced the amount of data that could be obtained without the user opting in, which - as privacy concerns grew - became the case less and less often.
Google didn't always help itself - the move from Universal Analytics to Google Analytics 4 wasn't handled very well, resulted in a very steep learning curve, and implemented modelled data without necessarily making that clear.
...we ended up here
Today, at a rough estimate, Google Analytics misses something between 25% and 55% of real traffic, depending on the audience - a tech-savvy audience will sit at the top end, a more general audience at the lower. end. But even 25% is quite a large blind spot.
This is compounded by Google applying statistical modelling to fill in gaps where data is blocked, or consent has not been given. But modelling works best when you have a lot of data - large sites, with millions of visitors, gives them enough data to make reasonable estimates. But smaller sites with a few thousand visitors do not, which means that the modelled figures are essentially educated guesses sitting on top of quite a thin data foundation.
Problematically, Google Analytics doesn't always make it obvious which numbers are real, and which are interpolated. So if you're looking at your Google Analytics dashboard, you may be looking at a number which is partly real and partly statistical inference. But without any clear indication which.
Does it matter?
In short, yes. We wrote an entire blog post on this top recently - Measure First, Market Second - which argued that marketing without measuring is like throwing darts blindfolded. You may hit something, you may not, and if you do, you won't know what you've hit anyway.
And this is a problem which is only going to get worse, not better.
So...what to do?
Don't rely on it entirely. You have other first party sources of data - if you're transacting online and pushing conversion events through Google Analytics, if it's telling you you've had 20 sales in a day but Shopify, Woocommerce or Magento is telling you've had 30, you know who to believe.
And consider having something in place running along side it that generates more accurate data. You may well still want, or need, to keep Analytics in place - Google Ads tracking depends on the Google Analytics / GTag being in place. You may have historical data that you need continuity for. It's often incorporate into other tools - Looker Studio or Search Console, for example.
But that doesn't mean you cannot find an alternative to have in place.
We practice what we preach
We've already installed Plausible to monitor traffic on our site. It's anonymous, so there are no GDPR / cookie issues. It's not an ad script, so it's blocked by ad-blockers. It's easy to set up. It's self-hosted, so web browsers won't block it.
There are numerous similar alternatives out there, some paid, some free - Fathom, Matomo, Cloudflare Web Analytics, Cabin - you name it - we just happened to plump for that one.
Conclusions
It's too early to tell - but after a month of side-by-side comparison we'll be reporting back on how much the figures differ.