Directed Edge, recommendations all the way from Berlin

Every once in a while we publish an interview with a start-up. We ask five questions, hoping the answers will give you inspiration and new views.

This time we’re interviewing Scott Wheeler and Valentin Hussong, co-founders of Berlin-based start-up Directed Edge. They provide web sites with a recommender system. Basically they serve three different types of sites, namely social networks, stores, and informational sites. Wheller and Hussong are now ready to conquer the world with their products and currently looking for a web applications and infrastructure expert and a marketing kinda guy. Exciting times. How did they get to this stage?

How did you come up with the idea of Directed Edge?

“Many years back Scott was doing some work on desktop search. Desktop search tends to not produce nearly as relevant results as web search algorithms because it’s basically using technology from the mid-90s. The critical moment in the modernization of web search was a shift from looking at text and tags and stuff like that to looking at how information is connected. So he was really interested in figuring out how to model connected data and use that to find related content.

If we fast-foward a few years, there was this “ah-ha” moment after watching movies suggested by IMDB as related to Scott’s favorite movie and being blown away by how horrible the recommendations were. So the pieces started dropping into place — recommendations are a huge part of modern web sites, combined with the rise of social media, there was a chance to use the social web’s structure to find content users are interested in. We started pitching the idea of doing recommendations as a service to some of our friends in the local web community and they helped us — and continue to help us — refine the idea.”

What was your biggest challenge during the development process?

“Focusing. I think focusing is something that’s a constant challenge for startups, especially if they’re started by people from a technical background. Technical people tend to look at a piece of technology and see what it could do rather than what problem that it solves. So there’s this iterative process where you think you’re actually focused, and that lasts for a month or two, then you realized that you need to zoom in more, then more and so on. We’ve still got a couple of iterations there. There’s a constant push and pull of seeing new places that recommendations can make the web better with trying to solve problems that businesses quickly identify with.

So, how can that help new startups? Well, I think it’s good to have the grand vision — that becomes your long term plan, but at same time you have to figure out where you can get a foothold at first and grow from there.”

Can you describe Berlin’s start-up culture compared to Silicon Valley?

”So, first, the good — Berlin has a great ratio of startup activity to cost of living. If you’re looking at trying to get something off the ground and keep the burn-rate low, it’s a good place to be. Berlin is also something of an Open Source mecca, so if you’re jacked into that community as well, there’s a lot of great hackers around.

On the other hand, the center of gravity in the German web scene tends to be in ideas that have already worked elsewhere, notably in the US. If you’re trying to create a new market you’re working up-hill. Most of the entrepreneurial and investment experience is concentrated there, and investors especially, and understandably, tend to stick to markets that they know rather than being wanting to gamble on new technologies.

There is, however, a new generation of German startups on the horizon, and there are a handful of projects that we’re really excited about, so we may see a gradual shift over the next few years towards a more diverse startup ecosystem.”

What will be the influence of your start-up on the next web?

“We’re part of the solution to “the relevance problem”. As sites grow, as the social web grows, there’s this explosion of content. There’s more information at our fingertips today than at any point in history. But in parallel sites struggle to show their users the right content — in a lot of cases that’s essential to keep them coming back, to sell more products and so on. So when we’re talking to a potential partner the question going through our heads is, “Where is relevance of content a pain point for these guys?” So, if you’re looking at local news, how do you show people the stories they’re interested in? If you’re selling tickets for concerts, how do you show users the concerts they don’t want to miss? Really, it’s about reducing the noise of the web.”

You can make up this question yourself!

What’s wrong with web recommendations today?

“Well, aside from often being just bad, there are two foci that we’re trying to sort out, that turn out to be related:

The first is that they’re too slow moving. They’re usually based on purchase histories or ratings or click-counting, so it takes a while before a site can recognize where relationships exist. Our approach allows information to move more quickly through the recommendations system so that our partners can help their users stay on top of the newest breaking stuff.

A side effect of those sorts of systems is that they tend to tell you about stuff that you already know about. Results tend to be too clustered. We’re more concerned about helping you find your new favorite band than reminding you that you still haven’t bought the 17th Beatles Greatest Hits collection. So we’re working on tuning our engine to find the real gems in a mountain of data rather than just letting you know what’s most popular in the genre where you’ve bought stuff in the past.”