Good news: it seems that Google Analytics has added a feature called “weighted sorts” to their ever increasing bag of tools they give us for free.
Why is this a good thing? There are probably lots of pages (in many cases, the majority), that aren’t in the top 100 — but, collectively, these pages may do a lot of work on the site. This helps you compare those pages, in real terms to the pages that are staggeringly popular; it makes smaller data sets comparable to your pages that have many more visits.
This gives you the opportunity to, if nothing else, understand the value of your segments, pages, search terms, et cetera using similar numerical terms. Just as segments are the lens through which you should be examining your data, this is now an indispensable lens to have in your toolbox.
Either way, it’s not an exact science, but it’s better than what you have now, which is nothing. And, because of how Google Analytics works, it’s not something that you could do yourself (because you don’t have access to every discrete visit which would allow you to weight them).
This is a really great post on the subject that gets into the meat and potatoes of it all.
There is one issue with the post, however, in that it overemphasizes the value of $index re:weighted sorts. Yes, it does show that certain pages on a user-per-user basis have a better $index. Those pages, however, have such a significantly lower amount of traffic, that your next step is questionable; do you take learnings from the better $index page with lower traffic and try and translate them to the lower $index page with higher traffic? Do you know what those learnings are? Does the lower trafficked, high $index page perform better because it is very appealing to a small segment that is, by its nature, small? How does that translate to the larger world of your site?
Anyway, it’s the perennial complaint with analytics which is “now that I see this data, what do I actually know? ” and then “what do I do with that knowledge?” That’s where analysis and testing come in. It’s the concert of data, analysis, and testing (and the iteration that you do with all three) that get you to knowledge that you can make actionable.
