Welcome Back, Mr. Lesbian!: Pitfalls of Perceived Personalization

 

Welcome Back, Mr. Lesbian!

In 1999, I created the personalization practice at barnesandnoble.com. About two weeks into the job, I got an email directly from the CEO… and I was being yelled at.

A customer had bought a book for his sister. The book was about lesbianism, and he bought it as a gift. At the time, the barnesandnoble.com home page had a drop-down box containing 50 different subjects, including history, fiction, adventure, sports, and gay & lesbian. When this guy returned to the site a week later, “Gay & Lesbian” was pre-selected in the box. From a company perspective, this level of personalization might seem reasonable: The user buys a book on lesbianism; ergo, the user has an interest in subject; ergo, highlight that section on the home page when user returns.

You’re likely thinking, “Yeah, but he bought the book as a gift. We already know how poor personalization is at figuring out what you bought for yourself and what you bought as a gift.” I agree with you, but that wasn’t the problem. What was the problem, then?

There was no personalization on the home page!

Huh? Then what happened? Well, when the code was written for the home page, someone thought it would be clever to put a random generator behind the list box. Whenever the home page loaded, the list box displayed a different subject as the “default” selection. One person would see “Fiction” highlighted, another would see “History,” and so on. On this day, on this page, on this particular refresh, the highlighted subject was “Gay & Lesbian.” The odds were about 1 in 50 any particular subject would be highlighted. That day, the odds were not in barnesandnoble.com’s favor. This man saw the category the site “suggested” and became irate. So irate he emailed customer service several times. How dare barnesandnoble.com make assumptions about his life based on one data point, especially when the data point wasn’t even relevant to him! The email messages went through customer service, up to the CEO, and down to me, with a “What the hell are you doing to our home page?!” tone. It was my second week on the job.

The home page had no personalization. But, it had a sense of perceived personalization. There was dynamic content that looked different to different users and could conceivably have been personalized.

The two principles of perceived personalization are:

1. People think you’re personalizing their experiences, especially if your site has dynamic content.
2. If you collect information that could be used for personalization but don’t actually personalize anything, your users will think you’re doing a bad job and get frustrated.

Let’s deconstruct these two rules. Rule 1 should be fairly obvious, given the above example. The lesson is many users don’t distinguish between dynamic content and personalized content. Content changes all the time, so why shouldn’t it change intelligently? If you have dynamic content that isn’t personalized, make it clear you aren’t selecting content only for that user. That way, the user won’t get upset because she thinks you are simply “wrong.”

Rule 2 is as much about privacy as personalization. Basically, you shouldn’t collect user information unless there’s a clearly defined use for it. In a previous column, I asked you to submit “my” sites you liked and disliked it was so I could write a “best and worst ‘my’ sites” column.

So We Can Better Assist You, Please Tell Us Your Social Security Number

I was doing research for that column and came across Topica (which one of you wrote me about). The site allows you to sign up for about a billion and one email newsletters. Great. I need more newsletters in my inbox every day. In its “sign up now” section, it asks for the basic information I expected (e.g., name and email address), then it asks for gender, birthday, and Zip Code. What?! Is Topica going to send me horoscopes? A Hallmark card on my birthday? If I don’t get a birthday card from Topica on my birthday (May 8, for those of you on expense accounts), I will be very upset. I wonder what it’ll do with the fact that I’m male. Perhaps the newsletters will be in shades of blue instead of pink. What about my Zip Code? Maybe it plans to send me weather reports. How thoughtful and considerate. Not.

In the eyes of the user, it asked for that information to enrich my customer experience, right? Nope. If you read the privacy policy, it uses “demographic” information for its market research. Topica can sell the information and prove its worth to sponsors. Oh, and these fields are required. What gets me even more upset is it is TRUSTe certified, which (in my opinion) means nothing.

Another reason a company might collect information is it “plans” to personalize user experience, but those projects haven’t yet launched. So, someone decided to go ahead with the registration first, collect the necessary information, then launch personalization sometime later. Bad idea. Wait until you have the entire system in place before you collect explicit information. Your users don’t know your launch schedule. They just know you have set expectations you’re not meeting.

In Conclusion

Avoid both sides of perceived personalization. If you aren’t personalizing your content, make sure your users know it, especially if you have dynamic content. On the other side of the spectrum, make sure you only collect information if you’re ready to use it. If you collect information for marketing purposes only, make sure the user knows that when you request the information. And make that information optional.

Do you have funny (or tragic) stories about perceived personalization? Let me know!

Until next time…

Jack

Jack is off this week. Today’s column ran earlier on ClickZ.