Under the page_pings heading in the event model, I am fascinated by the x and y offset events, showing where a user has scrolled since the last page ping.
But I struggle a bit with the best way to use this data analytically. On a user-by-user basis, I can step through and reconstruct where they are scrolling, perhaps, but what have others done toward, a more aggregate measure here?
For instance, is “total % of website scrolled to”, or “scrolling per seconds” things that others have computed? Any ideas would be great…it feels valuable, but also tricky to generalize (like a lot of the best data is, I realize.)
As long as it’s web events, no - the size of the viewport is compared to the size of the document. I think the tracker gets all of this info from the browser.
Hi there Andy. In my data model we basically end up with a scroll-depth percentage column for each pageview. It’s basically MAX(pp_offset) / MAX(pp_doc_height), so the deepest you go.
Then to analyse this, you probably want to categorise your pages into a page type. You’ll treat a home page or hub page differently from a content article page. Navigational pages might get deep scrolling, but it’s not a disaster if they don’t. Reading pages should get deep scrolling, though there’s exceptions too. We have a few articles on our site that are pretty quick references, so people come and go really quickly after getting the right piece of information.
With this kind of categorisation you can now compare things against their peer pieces of content to see the outliers. Ones that get lots of reading (think time spent too) vs things that get less engagement. Then you can start thinking why and rework the ones that aren’t working based on what you learn.
Of course if you have continuous scroll, your scroll depth is gonna be rubbish. And think about users changing orientation on mobile devices too!