Email Personalization
(This was the very first paid email newsletter I sent out, a little over a month ago. I'm publishing it here because people have asked for examples of the content I publish to my letter.ly. You can sign up for future newsletters at letter.ly/dave. It'll cost you $4/month — the price of a latte — but as a group you can help get a poor entrepreneur out of NJ.)
There's been some talk recently (Fred, Dave, Mark, Mike) about how email needs to be fixed. The general complaint is that recency is a terrible sorting algorithm and what would be better is a system that took into account the importance of the sender, the relevancy of topic, the closeness in social network, etc. Joshua thinks email is stuck because users are so used to email that they now reject any new innovations (which is an interesting concept in itself), rejecting startups like Gist, Xobni, Etacts, and Rapportive. Even Gmail (with it's marginal innovations) is a distant third to market leaders Hotmail and Yahoo.
So what's going on? Why can't we "sort by magic"? The problem isn't with the algorithm, it's with the user experience.
Getting a personalization algorithm right is really hard. You need to have massive amounts of data, strong signals of intent to sort through the data, and checks in place to avoid overrepresentation of popular items. Amazon can do it in books, but has trouble in apparel (seasonal items don't have any purchase data when they hit the shelves, and too many people go to amazon to buy "safe" apparel like white t-shirts and socks). But one of the most underrated necessities of a successful personalization feature is the user experience.
If you go to Your Store on amazon, you'll notice that each item that is recommended to you comes with an explanation as to why that product was chosen. "Recommended because you purchased X" or "Recommended because you added Y to your Shopping Cart". This is critical because algorithms are by nature black boxes, and people (who generally distrust technology) hate what they don't understand. So the first step is to explain why you are making the recommendation.
The second thing you'll notice is the "Fix This" link. This is just as important as the explanation, as it changes the algorithm from a hated black box to something the user can improve. Bought that gadget as a gift? Remove it from consideration. Want to see more variance in your results? You can change that (OK, not on Amazon, but theoretically).
So the way for email personalization to work is to explain why the algorithm thinks these emails are the most important ones for you to look at and provide mechanisms to tweak the algorithm so that it truly is perfect for you.