Information Management Weblog

Personalization

I've recently seen a spurt of activity denouncing the concept of personalization. News.com reported on a Jupiter Research report that concluded that personalization was "not only ineffective, but surprisingly costly." A few days later, Gerry McGovern jumped in with a similar rant about personalization and portals. Meanwhile, the folks on the SIGIA-L mailing list jumped in with a thread on the subject. Of course, all of these probably owe something to Jakob Nielsen's 1998 article declaring personalization "over-rated." And he probably wasn't the first to critique the idea.

Regardless, here is one of the better quotes from the McGovern article:

The best definition I have found for a portal is as follows: "A portal costs four times more to buy and operate than a normal website. It delivers half the benefits." Yes, that big, cumbersome, complex portal of your dreams may well make you long for the days when you were running a simple but effective HTML website.

The word "portal" seems to be one of those great buzzwords that pops up among those who know just enough to be really dangerous. I had thought the portal craze had run its course in the late-1990s, but I've been hearing it more and more lately. McGovern goes on to encourage people to focus on the basics: write good content, lay the pages out well, create easy-to-use navigation, and make sure search works. He's right of course. If you do the basics, then you can move on to the bells and whistles.

But, is "personalization" a bell and/or whistle you should add to your site?

First, let's define personalization. Simply, it is any time you serve up a customized page based on an individual user's profile. The classic example has always been Amazon.com, where the products and links you're shown vary depending on the pages you've viewed and products you've purchased. In in intranet situation, the site could show you different content based on your department or job type. Or, search results could be filtered based on your past search history.

There do seem to be some nifty advantages here. But there also seem be some big traps to avoid.

The navigation scheme should not be customized. A site should provide a consistent navigation scheme thoughout the site and across multiple visits to the site. A navigation scheme that jumps around depending on the users like likely to make users think too much.

There has been a fair amount of interest in personalized search. The concept has promise, but plenty of potholes, as well. Most implementations make use of a user's search history to influence results. Of course, each time a user turns to search, he or she likely has a different goal in mind. Even when using similar search terms, the user might have a different goal in mind. I doubt the additional cost and complexity would be worth it. I'd look at adding a taxonomy-driven "best bets" system before trying to personalize search.

Many e-commerce sites have "Other users bought these..." type features. These often lead users to other similar products. But often, especially if there isn't much data to feed into the Bayesian filter system that drives these features, the other items are somewhat random. The technology that drives these features is obviously getting better and better. But, right now, it takes some fairly seriously processing and data mining to put a system like this in play. Its probably out of reach for most sites. A better approach would be to make good use of metadata to display a similar items list.

So, are there good personalization features? Sometimes. Is it worth it? Nope. Looks like I'm piling on the bandwagon.