Towards Personalization: Strands Founder Francisco Martin (Part 3)

SM: Had you already built an initial version of the software by the time you received your first round of funding?

FM: At that point and time we had launched our first version of Recommender as an online service. We separated the Recommender engine from the final product since the very beginning. However, the best way to show off the Recommender is to tie it together with a consumer facing product. We built the music recommendation tools as a capability demonstration.

At that stage our goal was to court big companies. We felt they could be increasing their sales and help people find the right things in their catalogues using the Recommender. We actually launched Recommender in February of 2005.

SM: What were the capabilities of Recommender at the time? What was it recommending?

FM: At that time it was just music. We were collecting and correlating information which enabled us to make recommendations for millions of tracks. We spent a good deal of time on the patent work behind the ideas as well. Afterwards we put the engines to work on all types of projects and we saw that the approach we developed really works.

SM: What is the approach? Can you tell me more about how you do it?

FM: The same way that Google creates links between web pages. The intelligence that Google uses is human being intelligence. If I were to create a webpage and I made links to other pages then the page I created was created with human intelligence. Computer software can be used to gather intelligence and process it. The theoretical aspect is the value compositions. Google uses the page ranking system and we use a similar approach. We are also looking for links, but they are not hyperlinks. We are looking for links human beings dynamically create when the play music, when they watch movies, or when they purchase items. Anything that results in an identifiable sequence of human behavior is something we can track, analyze, and correlate in a large matrix.

SM: It is not collaborative filtering.

FM: If you define collaborative filtering in a wide sense, meaning that you are using the intelligence of many people in order to build a filter, then yes it is collaborative filtering. If you look at the implementation and the way we are generating the recommendations it becomes clear there what we have built is much more sophisticated. We have created a path that is very flexible to create recommendations which can be completely different. We started with music and moved very quickly to videos. Soon we will be moving into personal finance.

SM: Recommendations in terms of stock purchases?

FM: It’s scope is broad and covers new ways to save money and new ways to invest money.

SM: Financial markets are your next vertical?

FM: We are at a stage where we must focus on reaching the market. One method to do this was through social media which includes music, videos and all that stuff. That is why we started there. Finance seems to be another prime market.

SM: This service is all available through Strands.com, correct?

FM: So far yes. That was a recent announcement of ours. What you can see online now is a quick and dirty example we did in the past. For the last 6 or 7 months we have been engaged in re-doing everything and this is we will re-launch in the next couple of months. The personal finance recommendation capability is a result of this recent work.

This segment is part 3 in the series : Towards Personalization: Strands Founder Francisco Martin
1 2 3 4 5 6 7 8