Tuesday, August 23, 2005

Not all data is created equal

It's been a long time since I've fallen in love with a new web service as quickly as Pandora. I saw a blurb about the site on TechCrunch, and managed to grab one of Michael Arrington's first invites. Although I'm not a music junkie, it seemed intriguing - providing customized music streams based on computer analysis of songs you liked. To be honest, I thought that I'd be more interested in how they classified the songs than in what music they showed me.

Boy was I wrong! I quickly set up a station based on Madonna (my single favorite artist), and started listening. At first, it pulled a range of songs, and I'd say I really enjoyed about 30-40% of them. Still not bad, but nothing I couldn't find from the right radio station. However, after giving feedback on about 15 songs that I especially liked or disliked, that hit rate quickly increased to 70-80%! But more importantly, it wasn't just songs I already knew I liked - I found at least 3 or 4 artists I had never heard of, and that I really enjoyed. (For those who want to snicker about how out of the music scene I am, or about my musical tastes, the artists were "Katy Rose", "Alice Deejay", "Saint Etienne", and "Le Tigre".)

About the same time, I was thinking about trying out a new RSS reader site - Pluck. I went ahead and registered, but when the time came to enter my RSS feeds, I thought about manually re-entering the 60 or so feeds I have in Bloglines, and I completely gave up on the idea. I searched for a few minutes, but I didn't find any way that I could export a list of feeds that I have configured in Bloglines, and that really upset me. After all, it was MY data, I put that list together, why shouldn't I be able to export it out, and take it to another RSS reader?

This frustration was reflected in rule #5 on Charlie O'Donnell's 10 Steps to a Hugely Successful Web 2.0 Company - "Don't hold users against their will". As I was stewing about this cardinal Web 2.0 sin and rating more songs on Pandora, I realized that I have a double standard about my data. I don't expect Pandora to export my ratings that I give on their service. In fact, there's a lot of data out there that I generate, but I don't mind that I can't export. My buying history on Amazon for example, or even the recommendations that I give to books and movies I've seen before. When I gave up on Netflix, I had probably rated over 500 movies, but I never felt upset that they didn't let me export that.

My connections on LinkedIn and my newsfeeds on Bloglines, however, fall into a category that I do expect access to - even if I want to take it to a competitor. LinkedIn finally started allowing vCard exports recently, so I guess I'll lay off of them, but this duality made me start wondering - where is the line? Under what circumstances is it a given that you HAVE to give access to your user's data, and when is okay to keep it proprietary?

My conclusion was that it depends on whether the data was a one-party contribution, or the result of a "conversation". On bloglines, I found the feeds myself, and added them to my list. Since it was my own contribution, I want to be able to take it with me. However, with Netflix and now Pandora, they presented me with items that I responded to with a rating. Since this data was the result of interaction between the service and me, I don't feel bad about leaving it behind when I leave.

I'd love to hear others' thoughts on this - do you segregate your expectations for data accessibility depending on how it was created?

P.S. In case anyone else wants to try out Pandora, email me at llamaatinamedotcom. The service itself is great, and after emailing them with a new feature suggestion, I got an email back within hours saying that they were already working on it!