I was listened to the “reCAPTCHA and Duolingo: Luis von Ahn” episode of the How I Built This with Guy Raz podcast. reCAPTCHA struck me as a phenomenal business.

reCAPTCHA solved a problem - differentiating human and bot users on sign-up. The authentication service was free of cost or had minimal charges involved. Using their solution, reCAPTCHA charged firms to digitise text not digitalisable by OCR. Since reCAPTCHA offered a similar service for digitising but did not need to pay for the labour, they charged significantly less than their competitors. Finally, it took 2 people a short time to build reCAPTCHA. They were able to fix multiple problems at once and built a business with their solution.

reCAPTCHA is a phenomenal business. Here’s why.

Problem: Part 1

Websites wanted to block bots from creating multiple accounts. However, they found it hard to verify if a user was a human or a bot.

Solution: Part 1

To solve the above problem, Luis first created CAPTCHA. CAPTCHA classified website users as humans or bots using a puzzle. The puzzle involved identifying and typing in distorted letters presented in an image. For practical purposes, this puzzle was solvable only by a human. If the user solved the puzzle, the user was classified as a human; otherwise, the user was classified as a bot.

Problem: Part 2

In aggregate, people were spending a lot of time solving (slightly annoying) puzzles. Could this time be better spent solving real-world problems? What puzzle could solve real-world problems and simultaneously identify the user as a human?

Solution: Part 2

Answer: Text recognition not parsed by OCRs.

Consequently, Luis created reCAPTCHA. reCAPTCHA presented two text strings to a user. reCAPTCHA knew the text of one of the strings with great accuracy. The other string was unknown. Users entering the known string accurately were assumed to write the unknown one accurately too. Once enough people wrote the same text, it was taken as accurately digitised and the system moved on to the next word.

A phenomenal business

Many websites wanted an effective user verification system at zero cost. At the same time, a few organisations wanted to digitise scanned archives. For example, NYT scanned its archive since the 19th century, but did not have the text available digitally, or Google wanted to digitise all the world’s books. A system reliant on human digitisers was very expensive and would take a long time.

reCAPTCHA could digitise a year’s worth of text from the NYT archive in approximately a week. They charged USD 42k per year digitised and provided a highly accurate solution. This is opposed to human digitisers, who would take longer, be more expensive, and potentially more error-prone.

To create reCAPTCHA, 2 programmers developed the code. The solution was distributed by websites authenticating sign-ups. Websites were happy to use the system because it was free of charge, and was a relatively simple service to implement. For website operators, user authentication was one less problem to deal with.

Despite its annoyance, users solved the puzzle because it was short and it acted as a barrier to the user’s end goal. It was not the most convenient system, but a border that people needed to cross.

Customers, like the New York Times, were happy, because their text was digitised at record speed, cost, and accuracy. reCAPTCHA earned revenues by solving multiple problems at once. Once the code was written, distributed, and customers signed on, nothing could stop them. Many other website owners, and customers could use the same system.

reCAPTCHA is a great example of a business that is small, would most probably remain small, solve problems for multiple parties, and make money in the process. What an amazing business.

Update: I subsequently heard the ACQ2 episode with Luis von Ahn. It was similar to How I Built This episode but with a bit more focus on Duolingo. Would recommend you to listen to this episode too!