The problem with personalization
A few weeks ago, I was looking for a new watch. I went to Amazon, browsed, searched, refined, and finally decided on which watch I wanted—and I immediately purchased it. Since then, Amazon likes to show me more watches, every single time I visit.
But it’s not like I have a hobby of buying watches; I just bought a watch! I don’t want (or need) a new one! From my point of view, its an annoyance and, to Amazon, it’s pure wasted space. Amazon, long touted for having one of the best “recommendation engines” in the market, has failed miserably in this case.
If you’ve ever read the prospectus for a stock or mutual fund, you always run across these words (or similar): “Past performance is not indicative of future results.” Something like that holds true for personalization or recommendations: past behavior is not (always) indicative of future desires.
The fact remains that, at some point in the future, I might be interested in another watch, for myself or maybe as a gift. But the first time I visited Amazon after purchasing my watch, I was looking for some music; the next time, I was trying to find a book I wanted; the third time, I needed a shirt. Basing recommendations on my past searches did almost nothing to help with my current shopping.
I have a friend who runs a business where she monitors the internet for a number of legal clients; she needs to track certain issues, people, or companies on an ongoing basis. The problem is that the major search engines (Google, Yahoo!, etc.) like to use her past searches as a way of focusing their current results. She’s learned, through hard experience, to only use search engines with her browser cookies turned off, thus preventing the search engines from identifying her with her previous searches. It’s the most effective way she’s found to force the new, unseen, and therefore useful results to show up.
While using past behavior to attempt to guess current intent probably provides an overall better experience for a large number of users, it also has a severe negative impact on those power users who are highly directed and who could be extremely valuable to the search engines. I don’t have a proposal for a full solution, but I believe that it’s something that needs to be addressed for search to grow out of its infancy.
With regards to online shopping, I do have one complaint: why does no online shopping site allow me to search by size? I’m a big guy, and I would be thrilled (and spend a lot of money) if I could go to Amazon and request “all slacks with a 58″ waist” or “all long-sleeve button-down shirts in size 5X”. The information is there in the database; why force me to wade through page after page of useless results, checking each item individually to see if it’s available in my size? Likewise, if a size or item is out of stock, why does it show up in my search results? What possible benefit is there to show me an item that I can’t purchase? The only result is frustration and a seriously annoyed user.
I would've thought Amazon had gotten smarter about their recommendations, but it seems they haven't. I wish they would let you distinguish between something you're buying for yourself versus something you're buying as a gift. I don't like seeing recommendations based on items I purchased for my 76-year old mother. Makes a girl feel old before her time.
Yes, that's absolutely true. I use Amazon for a ton of purchases, and many of them are not for me. I know that companies are very big into "passive" personalization (i.e., it happen automatically without request), but there's a lot more power in "active" personalization (where the user can control what's happening). IMHO companies should provide both.
They have, but you have to manage that option. If you go to Your Account > Improve Recommendations, you can check items that were gifts and / or say "Don't use for recommendations." But really? Who wants to do all that work just so Amazon can sell you more stuff?