The Early Days of Web Analytics • Hacker News Board

Post by crdb » Tue, Jun 16, 2015, 5:23 AM UTC

exelius wrote: post #3

degenerate wrote: post #2

I'm using Analytics and occasionally tried other products that promised real-time data before Google caught up to them, but nothing stood out. I also trialed MouseFlow and ClickTale thinking that mouse tracking and watching individual users navigate my site was the way of the future... but the time required to setup each testing scenario, wait a few hours, and then watch a handful of users clumsily click around my si…

This is easy enough to do with GA if you use custom events. You don't even really need to set up much on the server side; if you fire the events at GA it will understand them. I think the real problem is expectations: Google is just doing statistics on big numbers, and for statistics to tell you anything, you have to have enough observations. The real problem is that you need a critical mass of customers -- millions…

Scale is a function of what you track, though. Taking e-commerce as an example: most people will track conversion rate (product purchases / product page views); most also track cart add rate, checkout rate, etc. But much fewer will track product impressions - that is, the event where a search returns the product in its results, or a category displays it in its catalog, and the user "eyeballs" it.

You get, rule of thumb, perhaps 1 conversion for 10 views; you also get 1 view for 10 impressions. So, tracking impressions (and therefore click-through-rate) decuples the amount of data you're tracking, and allows you to draw conclusions much faster than by waiting for the (admittedly, stronger) signal of conversion rate to tell you what you want to know about your product.

The real problems with Google are the ridiculous fees (150k USD/year for Premium? and I don't even get to talk to an actual Google employee if I have issues, have to go through reseller?) and the fact that there is a lot of "secret sauce" that you do not see and that isn't, or is sparsely documented (compare Google's last click attribution to what you're actually seeing appear on your raw data as the last click, for a simple example). Google wants you to use their superior insights; whether or not they are superior, I prefer having access to the full picture and draw my own conclusions.