Recommendations for me mixed up with recommendations for others
Ha, so much for smart recommendations. How do sites like Amazon handle recommendations when you spend a lot of your time shopping for other people?
I do most of my Christmas and birthday gift shopping on Amazon, so the recommendations it makes for me are based on things I buy mostly for other people. For example, I can say for sure that I do not want a book called Building Scalable Web Sites: Building, scaling, and optimizing the next generation of web applications, even though Amazon suggests that I might. Although, maybe that recommendation actually is useful, because someone I shop for fairly often might be interested in that book.
I rarely shop for clothing on Amazon, so I can’t speak to how useful its recommendations might be for my fashion shopping. But I would hate to get clothing recommendations for me mixed up with clothing recommendations for others. If I buy a sweater for a family member for Christmas, I doubt I’ll want to wear that sweater myself. I worry that such a purchase might confuse a recommendation engine. The recommendations it makes for me won’t be as useful if they are all mixed up with the recommendations it’s making based on items I bought for others. In fact, I generally ignore Amazon’s recommendations for just that reason.
So, is there a way to indicate that the item I’m buying is for someone else and it should not be considered when making recommendations for me? If there isn’t, there should be.
Jenn @ July 3, 2008
Yep, there’s a way to fix it. Go to your recommendations page. Each item has an explanation as to why it’s recommended, with an option to “fix this.” There you can remove it from consideration. I think this actually used to be one level up - on the top level of the recommendations page - in the past. But at least it’s still there.
I used to be a personal shopper and wished my most-used websites had some type of “folder” option where I could sort what had been ordered for whom.
PS: Am obsessed with that zebra scarf.
Great - thanks for the tip, Julie!
It does seem like there might be a more sophisticated way to manage recommendations. Barchbo, I really like the idea of folders and ways to classify recommendations for different purposes/people.
And I’m with you on the scarf. Now, if only it was November.
There’s also a place where you can set up a gift recipient list, and then look through your purchase history and assign the items to different recipients, or label them as not gifts. I’m not sure if this affects your personal recommendations, but it does help build recommendations for your recipients, which is pretty cool.
There’s a lot to be gained from a good suggestion-generating algorithm. For instance if you buy two things at the same time, they are probably for the same person, or at least for people at the same address. Amazon shows you “people who bought this also bought that” because it really makes sense to buy those things together. That works only because you are looking at one product that definitely associates well with another.
It can get very noisy if you only suggest based on past purchases. If 75% of what you suggest is appealing, I’d call that successful, but you need to show several options (if you’re going to show just one, it had better be good). Much like Amazon, you can rely on context of what they’re already looking for — put the peanut butter next to the jelly.
Of course, choosing clothing is much more complicated than choosing DVDs! If you can get data on a product’s size, measurements, dominant colors, etc. then you could build some background data about what a person likes. I consider this too hard unless the broad masses of data are really well organized. Microformats could make it happen, and then your suggestions could be effective. However I would still rely mostly on showing a lot of options, and basing it more on what they are looking at than what I think they like to buy.