Sramana: When you first recognized the concept of croudsourcing was working, what stood out at you the most? What about it make you recognize its overall potential?
Lukas Biewald: When you get thousands of people looking at your site or your data, you will get a lot of interesting ideas. They will point out confusing points that you may not have noticed. I reached back to my classmates back at school who were still grad students and were labeling tons of data themselves, and I showed them the crowdsourcing concept. They immediately recognized that it was a better way to get the job done.
Right around that time Mechanical Turk was starting. I then went to a lot of the companies in the Powerset investor’s portfolios and told them that they should be using Mechanical Turk. I spent a year taking a lot of Powerset’s processes and turning them into crowdsourced processes. The more and more I helped Powerset and other companies develop crowdsourced processes I realized that there was a company to be built. I kept setting the same people up to manage the quality and the throughput.
Sramana: When was this?
Lukas Biewald: It was around the end of 2007 when I left Powerset with Chris and set out to start CrowdFlower. We started a company that would make crowdsourcing a real tool for enterprises. Big companies could really benefit because they have so much data that they need to process and it is very hard for them to predict the volumes that they need. This was all happening before crowdsourcing was a word.
Sramana: What was your plan in terms of a business model?
Lukas Biewald: When we started the business the concept was that Chris could do the front end and I would do the backend and we would create a whole new industry. The truth is that we needed a salesperson. It took us a long time to get an early investment. We spent a year figuring out what our sales pitch would be. It was not obvious back then that scalability was what big companies would view as the most important aspect. At the time we definitely looked like a two person company.
Sramana: That is really not a problem these days.
Lukas Biewald: It is not a problem with Silicon Valley tech companies. It is not impossible but it is not trivial. I can go into an innovative company like eBay and essentially what I am asking them to do is take an important process that they care about and send it to the Internet to be completed.
Sramana: You and Chris spent a year working on this application. What else was going on during that time? You guys were self-funded and the economy was collapsing.
Lukas Biewald: My parents were asking me if I had a job. Normally, the escalator to success is that you go to Stanford and then you get a good job. Suddenly I had no credentials. It is great that Silicon Valley is so supportive; otherwise, we never would have made it.
Sramana: You built the application. Did you go out at get early customers even though you were not salespeople?
Lukas Biewald: Yes. That was when we learned the biggest entrepreneurial lesson – you have to be able to sell! We started with startups because they were the only ones willing to do it. Scribd was one of our first customers. We went to their office and they had three people and we had two people. It was pretty funny. However, they really needed us. They had to categorize all of the content that users were uploading. We helped them do that.
Sramana: What was the size of that project?
Lukas Biewald: It was millions of documents. Today we have a little over a million people doing work, but back then we had 10,000 people doing the work.
Sramana: How much do you pay them?
Lukas Biewald: It depends on the task. Categorizing a document may be around five cents per task. Some of our customers want us to go to a business and take a picture of it. That might be 10 dollars up to 70 dollars per task. The concept is to execute simple, short tasks.