The Learning Curve

New tricks for an old dog.

Trend spotting, correlations, or useless data

“Visitor screen resolution is a perfect example of a distracting metric that adds little value in any scenario . . . the metric rarely changes more than once every six months. Yet we keep reporting on it all the time, causing both a distraction and a sub-optimal use of time. Besides, would it not be a better strategy to simply use research from Forrester Research or Gartner on the latest trends in your industry segment and use that to drive the size of your web pages should be?”

Avinash Kaushik, Web Analytics: An Hour a Day, page 9

This has all the earmarks of being a pet peeve. Try as I might, I have a hard time imagining that including screen resolution data in web analytic reports is such a huge distraction or waste of time, so as to justify the time and effort involved in giving it the ax.

Web sites capture a collection of data for each visitor, but every analytic program I’ve ever looked at (which isn’t all that many) immediately splits the data apart and aggregates it by type. So, for any given period of time, you know how many and what percentages of your site’s visitors are using what screen resolutions, including pretty little pie charts.

Now, wouldn’t it be interesting to see if there is any correlation between visitor screen resolution and newsletter sign-ups? I’m not aware that anyone has ever bothered to slice the data quite that way before, and the reason why they haven’t done so is clear.

It doesn’t make any sense.

It’s difficult to imagine.

How could there ever be a connection between screen resolution and newsletter sign-ups?

Predicting human behavior based upon an assumption of rationality is probably the biggest mistake anyone could ever possibly make. This is not to say that people are totally nuts, even if quite a few are. People are rational in a way that makes sense to them, but that does not mean that their rationales or behaviors will make sense to you.

That’s the whole justification for metrics and research in the first place. If everyone thought and acted just like you, you wouldn’t need to ask because you’d already know.

If you don’t look at the correlations, you don’t know. You can fool yourself into believing that you do know, or that it isn’t important, but nevertheless, the fact of the matter is that you just don’t know.

It might be important.


Written by Tom Fox

09/21/2007 at 8:30 am

Posted in Learning, Web Analytics

%d bloggers like this: