Web analytics are highly susceptible to error, manipulation and misinterpretation. They should never be depended on as the sole source of insight.

“Facebook might be hosting upwards of 8 billion views per day on its platform, but a wide majority of that viewership is happening in silence. As much as 85 percent of video views happen with the sound off, according to multiple publishers,” Digiday reported in 2016.

The conclusion many people made when they heard these statistics was that they needed captions for their videos. I was a bit more skeptical. Having dealt with Web metrics since the late Nineties I had come to approach with them with a lorryload of salt. My first question was: Are most video ‘views’ real?

Facebook just settled a legal complaint which stated that, “The average viewership metrics were not inflated by only 60%–80%; they were inflated by some 150 to 900%.” That’s a lot of inflation. These fake statistics had real consequences.

“In order to beat YouTube, Facebook faked incredible viewership numbers, so CollegeHumor pivoted to Facebook,” Adam Conover states. “So did Funny or Die, many others. The result: A once-thriving online comedy industry was decimated.” Of course, these fake statistics were no accident. “You’ve had 18 measurement errors in recent years, all of which went your way,” Terence Kawaja explained to a Facebook ad business ethics executive. “That’s not a mistake, it’s a strategy.”

Web analytics in general are incredibly flaky. In the early days, we had HITS (How Idiots Track Success), which were an absolutely useless metric but were quoted widely because HITS were the BIGGEST number in the metrics report. This reflected the culture of metrics and the Web in general: chasing BIG numbers. I’ve had people tell me that they can’t delete out-of-date and incorrect content because it would reduce the number of visitors to their site. Others refuse to make their sites simpler because they’re wedded to metrics about getting people to spend lots of time on the site, clicking on links and scrolling and stuff. The fact that, on a badly organized site, time spent is more a measure of failure is rarely recognized in the mania for BIG numbers.

Today, most websites don’t have much of a clue how many people visit anyway. You may have lots of ‘users’ but the same person using your site on a smartphone and then a laptop is likely two users in your stats. Recently, Matt Hobbs, head of frontend development for GOV.UK, explained that there are 50–55 million user visits a month to GOV.UK. But they don’t know how many citizens this represents.

A great many digital teams are unaware of what is actually happening on their website or app. This is one more reason why you need to observe your customers regularly. It gives you real-world context through which you can understand your Web analytics, and it helps you identify where flaws or misinformation may be occurring. Regular observation of real customers doing real things on your website or app is so essential. It gives you the crucial context of the reality of use.