Even more on why I think time spent on site is kind of a dumb metric
I don’t mean to harp, but “Average Time on Site” is a silly metric to use. Not in theory, sure if you could actually calculate it and know the length of time people had your site open and were reading it, it’d be the best metric ever. Unfortunately, you can’t.
So that the number is a lie. Sure hits, visits and visitors aren’t completely precise either, but those are much more calculable and the numbers you get as you refine are much closer to the truth, since it’s a very specific and known thing you are trying to calculate. The server knows how many times it has served a page and it’s a matter of weeding out the ones you don’t want to count. With time spent on site, no one knows who’s paying attention to your site, know one knows if the person clicked around your site, took a smoke break and came back 20 minutes later to finish or if he looked at one page and reread it 20 times to take it all in. It is unknowable.
So how do the numbers makers deal with this? They just make things up. Don’t believe me? Check out this post on the Official Google Analytics Blog where they say due to customer pressure, they’re changing the way they calculate the number. That’s right they changed it not because they learned some new thing that would make them more accurate, they changed it because people complained.
How often does Google cave in to customer pressure? How often does Google care what their customers think? They do what they do and you thank them for it. If they were weak kneed they couldn’t be in the search engine business, they make their algorithms and they stand by them in the face of whithering critique. Except for this one, because they know it is just an arbitrary decision - neither number reflects any sort of accuracy, so who cares which it is? As more companies pay more attention to time spent on site, the way to calculate it will become more homogenized and various analytics will begin to show largely the same numbers - this will give the illusion that it is more accurate, but in fact it’s just as inaccurate as it is now but everyone will be in line with the same lies.
And this one change they made, simply changing how they counted one type of visitor (the ones that come in for one page) dropped the number by like 90%. That’s right. And that’s the nature of time spent on site, there’s no right way to do it, it’s a series of arbitrary decisions, the changing of which drastically affects the resulting metric - affects it significantly more than changes in the actual data would. In other words, how much would my readers’ behaviour have had to have changed to affect the same 90% difference?
The best part is that everyone in the industry knows this, but they all want to have a number and if they all agree to believe the same lie, then at least they’re all playing on (fictional) level ground.
