categories

HOT TOPICS

NEWSLETTER

If you are considering becoming a 1M/1M premium member and would like to join our mailing list to receive ongoing information, please sign up here.

Subscribe to our Feed

CrowdSourcing Repetitive Tasks: CrowdFlower Founder Lukas Biewald (Part 3)

Posted on Saturday, Oct 1st 2011

Sramana: What is your business model?

Lukas Biewald: Our business model is pretty simple. We quote customers the price for completed tasks, and upon completing those tasks recieve payment regardless of the cost we incur competing those tasks. We assume the price risk. Let’s suppose a company has a large directory of other businesses and they want to provide those addresses to their sales force. However, before sending salespeople to those addresses they want to validate that those are the correct addresses. Convenience stores are a good example because they close and open new ones all the time. You don’t want to send a salesperson to a store that has been shut down.

That company would then have a task of checking all of the addresses they have in their directory. A project that we work for them would then turn into us checking a million addresses for them at the cost of 1 million dollars. I am making those numbers up, but it conveys the concept. The first thing we will do when we receive the task is to break it up into pieces. The first step would be to find the website for the business. The second person would check the work conducted by the first person. If they can’t find the business online then a third person could be tasked to call the business. If that does not work out then a fourth person could then be tasked to physically go see the business. Each task is increasingly expensive.

Anyone can put jobs online. The CrowdFlower technology that actually makes this work for big companies is the quality control. We keep track of people. We recognize people that have done good work in the past and give them access to higher paying jobs, but we will still watch and check up on you 10% of the time to make sure you are still doing a good job.

Sramana: How do you do those quality checks?

Lukas Biewald: We will have a second person check the first person’s work. When you do it at the scale that we do it at, the second person might get it wrong as well so we will bring on a third person to check the second person sometimes. This is where my AI background really fits in well. There is a ton of math to make this platform work because of all the probability. The customer may demand 99% accuracy and we have people with 95% accuracy rates, so we need results validated somehow to achieve that 99% accuracy.

Sramana: Essentially CrowdFlower receives tasks from the customer at a fixed price?

Lukas Biewald: We receive it at a fixed per-unit price. It is a lot like cloud computing.

Sramana: So corporations pay you and you are able to leverage a crowdsourced workforce around the world that executes the work. You simply have different rates for the tasks to get the work done at a lower price and you earn money on the differential. Is that correct?

Lukas Biewald: Yes, you broke it down very accurately.

Sramana: What do you think your margins are? If you got a 10,000-dollar project how much would you make on it?

Lukas Biewald: It really depends. We do enterprise sales so our margins vary a lot depending on how mature that application is. We will make better margins on jobs that we have done a lot. We will also make better margins if we have people come in and they are more accurate. If we find more URLs, that is really good for us. If the customer gives us a bunch of convenience stores that happen to be online then we are going to be able to check them online, but if we have to call every store then we may end up making negative margins. I almost think of it as Google CPA pricing where Google takes all the risk in exchange for an unknown margin on any give task. That keeps things interesting.

This segment is part 3 in the series : CrowdSourcing Repetitive Tasks: CrowdFlower Founder Lukas Biewald
1 2 3 4 5 6 7

Hacker News
() Comments

Featured Videos