Lessons from Findory

Greg Linden is putting Findory to sleep:

Development on Findory now will slow to a crawl. There may be new features, but they will be rare. I no longer will spend time exploring funding, biz dev deals, or recruiting.

Findory and outbrain share the same goal - to help people find content that's potentially most interesting to each one of them (as opposed to most-popular items as in Digg, etc). Findory was based on the same collaborative filtering concepts behind Amazon's personalized recommendations, which was also authored by Greg.

So what went wrong?

Why did the algorithms that worked so spectacularly well for Amazon's personalization didn't take off for Findory's personalization? I think this happened for one core reason:

• In shopping: personalization = helping me find that one item that I'm likely to buy and love. Or in other words, personalization via recommendations.
• In content: personalization = help me save time by sorting through the piles of news so that I can skip through the bad stuff and spend time only on the good stuff. Or - personalization via filtering.

I bet many people (me included) would love to sit around and consume content on Findory (or Digg, etc), but just don't have the time to do so...  Findory required me to spend more of my scarce attention to use it. What I need is a service that has 'net positive attention emissions'... A service that saves me time rather than consumes more of it. We're not there yet, but that's exactly what we're trying to do at outbrain.

Greg is one of our favorite bloggers at outbrain, and I hope that the fading of Findory doesn't also mean the fading of Greg's blogging.

More coverage on GigaOm, TechCrunch, Don Dodge and Read/Write Web.

Comments

[Ed Billingsgate]

I'm all for an easier way to find interesting RSS feeds or blog posts that would jive with what I like to read, and I'm excited about what Outbrain is trying to do -- it is a huge step to take. I hope they're not going to try to re-invent the wheel, though, as a collaborative filtering engine is not something you just write from scratch (not one that is consistent and easy to maintain as your traffic scales). I'm hoping they can utilize a service such as CRITEO to outsource their collaborative matching needs to someone already working on offering a solution to that problem. If they do put that burden on someone else, hopefully they can work on marketing their idea to the blogosphere so that others can interact with the system, rather than offering a very closed-system that requires their own labor to get the momentum going.

I actually think that this idea of outbrain's is one that many are working on, but I think most will get stuck on the two biggest issues: (1) how to handle the actual collaboration engine, and (2) how to handle adding blogs to the engine and allowing users to tag or vote for what they like and dislike. In my opinion, both of these huge limitations can be resolved by outsourcing at least the collaboration engine.

Are you involved with outbrain at all?

[Yaron Galai]

Ed - Thanks for the thoughtful comment.

1st - I am a founder of outbrain... I apologize if that was not properly disclosed in the post above.

2nd - Criteo has done a great job on their collaborative filtering platform, but again - that's an example of 'filtering via recommendation'. I'm not sure how immediately applicable and useful this would be in the world of news and RSS. There are so many nuanced differences between those two worlds (consumption of products and consumption of content), that I think eventually we'll have different leaders emerge in each space, each specializing in their specific filtering task.

[Ed Billingsgate]

Yaron -- Thanks for the clarification. I just stumbled here from a Google search, but I'm a subscriber now :)

As for what Criteo is doing, you might be right but it is an intriguing thought that I'd like to see someone working on. What I noticed about Criteo is how they allow for topics (such as in the Movie matching demonstration by type), which would be similar to tagging a blog/RSS feed. By utilizing some sort of global tagging system, it should actually work fairly well to use a recommendation engine to perform a collaborative filter across a variety of RSS feeds of blogs that might not be topically similar but specific posts within those blogs might (based, again, on the tagging by either the blog author or by visitors who utilize whatever collaborative engine is used).

I'll be watching outbrain to see what their ideas are, I just thought I'd share what I've considered personally.

Thanks for replying!

[Yaron Galai]

Thanks for the ideas, and thanks for joining as a subscriber!

I agree there's lots of value in looking at tags (or other meta-data) and gaining intelligence from that to support the pure collaborative filtering. The question we were debating internally is whether we'd want to develop stuff like that internally and compete with existing services, or whether we'd focus on being good at what we set out to do and partner with others to pull in data like that. Where possible, I'd prefer to opt for the 2nd route.

Again - thanks for sharing your thoughts here. Keep those coming!