Wednesday, May 4, 2011

Design by the wrong numbers

I'm presently beset with a question, the same question with which I've struggled for many years: what is the average amount of time a person will spend on a Web page? The question seems to arise at the beginning of most projects, and it's a question to which there has, to date, been no satisfactory answer. But what's struck me about it this time isn't the lack of a reliable answer, but that it is likely the wrong question even to ask.

One problem is that it is backwards-looking: how much time a person spent on the current page is not necessarily a good indication of how much time they will spend once the page is redesigned. Different content will get a different level of attention. A person may spend mere seconds on a page of search results, and several minutes reading an article. To pages on the same site may garner different levels of attention. There can be no single number that suits all pages.

A second problem is the amount of time a user chooses to spend perusing a page is something that can be controlled to effect a positive outcome: there are various tactics that can be demonstrated to affect the amount of time a user spends on a page - more or less - but what's lacking is the consideration of whether that has a positive or negative impact on the goals of the site operator. Again, you can examine historical data to determine that a user who spends a given number of seconds on a product page is more likely than others to buy - but this is likely circular logic: a person who is more serious in their intent to purchase will spend more time than a person who is not - so statistics of time spent and purchase likelihood correlate to interest and not to one another.

A third problem is that statistics is based on the assumption of a bell curve, with most measurements clumping toward an average and diminishing toward the extremes. Observations of behavior on the Web tend to display more of a U-shaped curve, with greater numbers at the extremes and fewer closer to the average, such that designing for the "average" will not have good results - and will more often be counterproductive.

That last bit might need a bit more explanation, as those who put their faith in numbers and eschew observation may find it hard to accept. And so, a quick analogy: examining the age of visitors to a given public park, one could accurately conclude that the mathematical average age of visitors is 29, whereas any causal observer could note that most visitors are children between the around the age of ten accompanied by grandparents around the age of sixty - and virtually no-one of the mathematically average age is present. Thus, redesigning the park to suit the average would lead you to tear out the playground and the benches and replace them with a running track and a soccer field - and as a result, the park would no longer be appealing to its current users, not attract a sufficient number of users of the "average" age to replace them.

The same is true for the visitors to a Web page: they are skewed to the extremes. The statistic that comes to mind is that the average time spent on a single page is around one minute, or sufficient time to read 200 words of content. Even if you exclude "spiders" and "robots" who create the impression of a very short visit, and users who are distracted and leave the page on-screen while they are doing other things, there will still be considerable concentration at the extremes: some users click to a page, recognize very quickly it's not what they wanted, and leave after reading the first few sentences; others have a genuine interest in the information and will read the entire content.

As such, designing Web pages according to the average amount of time spent will result in bad design - and given the second point, that users who are interested in buying will likely spend a long time perusing - the design will be skewed to suit the least valuable audience for the site operator - in effect, to provide enough information to enable a non-buyer to leave, but not enough to encourage a likely buyer to continue.

No comments:

Post a Comment