I just followed the trail from John Battelle‘s note about Yahoo’s launch of Mindset to Greg Linden‘s post comparing it to Google Personalized Search and MSN Search, then to Greg’s company Findory.com and his history at Amazon helping develop recommendations and personalization features there.
Personalization of search results, news and blog stories, product recommendations, or anything else depends on “stuff about the user,” namely preferences and interests. Different services take different approaches, but they all have to grapple with two questions, each of which has three possible answers:
(1) How will the user’s data be obtained?
(A) by asking
(B) by tracking behavior
(C) by matching the user to separately collected data
(2) Where will this data be stored?
(A) it won’t, it will only apply to one action
(B) in a cookie on the user’s computer
(C) in a user account set up at the site
Let’s check out the approaches taken by the above services:
– Amazon lets you create an account that is automatically populated as you buy or rate products (1A, 1B, 2C)
– Findory lets you create an account which is automatically populated as you click on articles and blog entries or search using the site (1B, 2C)
– Most search engines let you specify search options either per search (1A, 2A), or as “saved” preferences, which are saved as a cookie on your computer (1A, 2B)
– MSN Search automatically fills in your location by extracting it from your IP address, and saves it as a cookie on your computer (1C, 2B)
– Google Personalized Search lets you select general interest areas and saves them as a cookie on your computer (1A, 2B)
– Several search engines offer further personalization features if you create an account with a search history (1B, 2C)
An example of how to avoid both questions is Yahoo Mindset, which lets you filter search results by using a slider that governs the influence of two preset preferences, whether you are “shopping” or “researching.” Thus no data is obtained or stored, but the personalization is very limited.
There seems to be no doubt that knowledge of user data can enable improved services, but the options above lead to several concerns, at least for me:
– If my behavior is tracked or my identity is matched to any other data, I want to know about it
– I always want to be able to view, modify, and/or delete any information associated with me
– If my data is stored in a cookie, it has less usefulness, since it only then works for that computer and browser, and may be deleted (accidentally or otherwise)
So I’m looking for personalized services that store my data in a user account, let me know explicitly how data ends up there, and then let me view, modify, and/or delete any of that data at any time. This is why Amazon (and Findory, which adopts its methods) are to me the most useful and comfortable to use.
But even this solution leads to some problems. If my data is stored in an account at the site, then:
– The data can only be used to help personalize services at that site; all the work I put into training Amazon won’t help Findory or anyone else
– Different sites offering similar services aren’t really competing; the first to get to me (Amazon) “owns” my data, and I can’t transfer, synch, or share this data with other services that might use it more effectively, I have to start from scratch
– I have yet another username/password to remember for each personalized service
To me, the essential fact is that I should “own” and control my own data; if my preferences (or purchase history, email address, etc.) are requested by a site for personalization or any other service that I value, I’d like to be able to exchange my data for this service (in some way that I can modify or revoke).
This reminds me of the “infomediary” concept John Hagel introduced in the book Net Worth back in 1999. But whether or not a third entity mediates the exchange, such a functionality would lead to a more level playing field and, I think, an increase in innovation associated with personalized services.