In Search of the Mythical Step Function

By on in , with One Comment

One thing I’ve learned over the years is that the real world of data is a lot messier than a whole range of information outlets imply. Whether it’s web data, CRM data, or ERP data, it is very, very seldom that there is something going on that, once discovered, can have an immediate and dramatic positive impact with little effort. The reality is, it takes some up front discipline to prepare to conduct an analysis, then, often, a not insignificant effort to get the data needed for the analysis pulled and prepped. And, at the end of the day, the best results pass the hurdle of “statistically significant.” What that means is that a set of variables may be found that have a slight-but-real impact on something you care about. Now, hopefully, those are variables that you can influence, and that you can influence without too much investment.

A classic examples with web data is what I call “the myth that people are cows.” Anyone who has ever been in a pasture with a herd of cattle knows that it takes the average bovine somewhere between 1 and 3 nanoseconds to settle into an unwavering pattern. Grove of trees to watering hole at 8:07 AM. Watering hole to grassy knoll at 1:53 PM. Grassy knoll back to grove of trees at 6:32 PM. More than that, the entire herd follows the exact same 12-inch wide path unerringly from point to point. It takes almost no time for that path to be a well-worn, dirt, 12-inch wide trail.

(For an absolutely wonderful poem on the subject of such paths, and how Sam Walter Foss imagined such a path driving urban development of a major city, check out “The Calf-Path.” at Public Radio International’s The Writer’s Almanac.)

The problem is that, all too often, Marketeers assume that people are like a herd of cattle. They know that, if they can just find the most common paths through their web site, they can take advantage of it in huge and profitable ways! “We’ll know exactly where to put a billboard for Maisy’s Magnificent Udder Moisterizer that will attract the most eyeballs!”

Unfortunately, visitors to web sites are not cows. Not even close. Try some simple math. How many unique links do you have on your home page? 10? 20? 100? I’d bet good money that, if you count them, you’ll realize it’s more than you thought. That’s the beauty of drop-down menus in that a clean and simple design can still present the visitor with a lot of options. For chuckles, let’s say there are only 10 links on each page on your site, including the home page. Let’s also say that the site does not have a search box that persists on every page (poor form, that). And, let’s go ahead and say visitors’ browsers don’t have a Back button. Given all of those unrealistic constraints, the number of unique possible paths from the main page of the site five levels deep is 10^5, or 100,000. With 100,000 options, the most popular path is going to be, at best a percentage point or two of the overall traffic. Now, factor in a Back button and a search box…and the math got wayyyyy to complicated for this blogger.

The point? It’s a waste of time to try to hone in on the “most popular paths from the main page of our site.” At best, it makes sense to look one or two levels deep…and the most likely insight you will ge there is that there is a noticeable chunk of people who are clicking on a link on your main page and then clicking the Back button, which should make you question if the page they clicked to is delivering what the link implies it will.

Web analytics vendors don’t really help things. Clickstreams are so popular among the underinformed that they have to bake clickstream functionality into their tools, and they need to do so in a way that makes for a slick demo. Of course, in the case of the demo, they have control over the data, so they can show a cow-like clickstream! Reality…just isn’t that simple!

I really can’t seem to right a short entry. I’ll try again next time!

One Comment

  1. Pingback Gilligan on Data by Tim Wilson » I Wonder If Came Up at eMetrics?

Leave your Comment

Notify me of followup comments via e-mail. You can also subscribe without commenting.

« »