Human vs Algorithm
We’re often asked about whether our Investability Rating system is human-driven or algorithmic.
Ignoring for a second the over-use of the word algorithmic, the answer is “both”.
Assessing the quality of a startup’s investor proposition is not a problem you can point a machine at. There isn’t enough hard data on your average startup; which is why startup assessment at scale is such a tough nut to crack.
Here’s our approach, and an illustration.
Remember Google CAPTCHA asking you as a security measure to click on images to identify those that had a bus or a bridge or a traffic light in them?
We never thought twice about how to answer those questions. We just clicked, and so Google got unbiased answers.
That’s our Investability Rating approach. Ask a capable human assessor a series of questions about a startup. Ensure those questions are highly specific and carefully phrased so that the assessor can just click. And you get unbiased answers.
We ask assessors lots of simple questions which can be answered easily, to break down big and difficult problems such as “how good is this startup’s team?”. Then, we aggregate and weight the individual responses and build them into an overall Investability Rating.
We have a proven system which promotes objectivity and reduces bias, and we now have lots of data to demonstrate that our ratings correlate with actual outcomes. A high rating is an accurate predictor of a positive outcome.
But what about scalability? How many human assessors do you need to have if your ambition is to rate many thousands of startups globally?
That’s where the other side of CAPTCHA comes into play. CAPTCHA was an important security tool; but its secondary purpose was to teach the machine how to recognise a bus or a bridge or a traffic light in an image. Google used humans to teach their machine.
We’re using our human assessors to teach our machine how to recognise an investable startup. It’s a more complex challenge, and we don’t expect to abolish human assessors any time soon, if ever. But our machine will bring automation and predictive capacity to many parts of startup assessment, ensuring increasing efficiency and scalability in our model.
And a freebie factoid: who knew that CAPTCHA is an acronym for Completely Automated Public Turing test for telling Computers and Humans Apart?