It’s that time a year again — my birthday. Every year on May 8, I sit back and reflect on the year gone by. I think about where I was a year ago and where I thought I’d be now. I think about what’s changed and what’s stayed the same. I think about what I needed last year versus what I need this year. I think about new music I listen to, different kinds of movies I watch, and new kinds of books I read.
Then, I go to all my personalized sites on the Web and update my accounts to make sure their understanding of who I am reflects all these changes.
It’s all true, except for that last part. I don’t update my Web profiles often at all. Not even once a year. So how can I expect a “personalized” experience from a site that doesn’t really know me anymore?
CDNOW or Never
In 1997, I worked at a start-up called Open Sesame. We built an artificial intelligence agent for use on the Web. Think Net Perceptions on steroids with better technology. For my birthday that year, my VP gave me a CDNOW gift certificate. It wasn’t the site where I normally bought CDs, so it was the first and last time I shopped there. Five years later, I still get email updates recommending CDs it thinks I’ll like. We all know one data point doesn’t define you (as I wrote in a previous article), but that isn’t the point. The point is that was 5 years ago!
I’ve changed, CDNOW. In fact, CDNOW, you’ve changed, too. Can we reconnect after all these years? Can we still be friends?
Five Clicks Versus Five Years: Which Defines You?
In my lectures, I discuss the difference between classical data mining and personalization. I always say personalization is the execution arm of data mining. Data mining technology crunches data and discovers patterns and knowledge about customers and your products. Personalization technology uses that data to affect users’ experiences in real time.
There’s another crucial difference. Personalization technologies make decisions in real time about which mined data is the most “true” at the moment. They take into account context and recent behavior, weighting these much more heavily than past knowledge. They have to “know” when certain data is no longer relevant and adapt over time as people’s interests change.
Back in my Open Sesame days, the technology we built learned over time. It would observe a user’s behavior and suggest how the user’s interests had probably changed in the recent past. If the user filled out a profile saying she was interested in country music, for instance, but the last three months of her browsing/purchasing behavior showed a strong interest in musical theatre, the system would raise a flag and say, “Hey — I know you said you liked country, but maybe now you’re more into musical theatre?”
Of course, it wouldn’t literally say this. It sent the system a sort of message. The Web site could decide what to do with the new knowledge: ignore the information, automatically update the user’s profile, or ask the user on the site or via a email if the noticed change was accurate and if system could update her profile automatically. This system could be user-controlled for privacy, so the site never learned your unspoken interest in certain subject, if that was your wish.
It’s simple: Interests change over time. A personalized site’s knowledge about you is probably stale if it’s over 8 to 12 months old. People don’t usually update profiles. The best system would allow users to explicitly say what they’re interested in, then offer to automatically update those interests over time if changes in habits are detected. That way, personalized sites would remain relevant and continue to provide the service they intended — presenting users with content and information relevant and current to their immediate needs and interests.
CDNOW should realize the information it has about me is stale and try to reactivate me as a customer. It should realize I’m not the same person I was five years ago.
So, happy birthday to me! As I think back on the year and realize how much I’ve changed, what my new needs and goals are, and what the coming year will be like, I wonder how many sites out there still think I’m the same person I was when I signed up, and how many will notice how different I am this year.
Until next time…
As an organisation, finding the right marketing channels is an essential part of your marketing strategy.
When measuring the effectiveness of discount codes, retailers often get it wrong. In this article, we'll look at how data-driven attribution can help businesses better understand where discount codes produce the best ROI.
Data. It’s the latest ‘buzzword’ in the digital marketing world when it comes to content.
Digital has quite forcefully overturned the entire media industry, causing even the most traditional companies to adapt or be left behind.