So - had a look at the YouNoodle startup predictor (we covered it
here, applied for the beta, got no response so just clicked on
via the TechCrunch link) - it is interesting in terms of where it spends its prediction resources:
- fairly high level analysis of the company structure, market you are in, and you can only have one sector so there are mutual exclusions - eg Internet and Computing for example
- extremely detailed analysis of the team, their experience, qualifications, relationships etc
The output comes in terms of a score out of 1000, and a valuation. Its hard to see how the score is made up, as there is no ability to analyse contributing factors - so for e.g. I don't know whether it would be higher impact to have an experienced advisor or another engineer who is better qualified for example. What was also interesting is it valued the company at roughly the value we put on it - but we told it first what we thought the business was worth, so one hopes it arrived at this independently.
I must say I thought some of these questions would probably breach some NDA's, and the personal data requirements were extremely detailed in some areas - and quite a bit of the personal data was not relevant to valuation but seemed more relevant to building up a funders' database.
Anyway, the other thought I had was I could see how it was valuing the company as a "black box", but I wondered how it was valuing the company within its ecosystem - that is:
- the 1st company doing X is valuable, the 10th is....well, it needs to have more to it.
- different ecosystems go in and out of fashion (the Gartner Hype curve)
- the overall market goes up and down - all our prediction models take account of sector and market sentiment.
I assume the model does this sort of activity deductively, but it was interesting it did not ask more specific questions (competitors identified, sectors we think we're in) etc - given its extreme interest in our personal networks its interesting it had so little interest in our value chain network.
Anyway, very educational process - we are certain this sort of analysis will be used more going forward, but would I really trust my company's detailed data to a small Webservice startup that could be acquired by who knows who? I'm not sure - would you? (For the record, we used an anonymised case study)
Update - on the sentiment value adjustmet issue, I saw this on
VentureBeat
To get a better sense of YouNoodle’s accuracy, TechCrunch asked the company to calculate a few more valuations of high-profile companies based on available information about their early stages. Goodson sent the analysis along to me. In cases where real-world numbers are available for comparison, YouNoodle was usually a bit off. For example, it predicted a $124 million valuation for social application startup Slide, and a $71 million valuation for competitor RockYou. As of their most recent fundings, the companies were actually valued $550 million and $250 million, respectively. That’s pretty far off, but Goodson notes that YouNoodle did predict that Slide’s founding team was stronger.
This tells me that YouNoodle is probably looking at company worth without huge adjustment for market sector sentiment (the Slide/Rock You sector being quite high-hype right now) - so it's great if you are getting valuations in offf-the boil technology, not so good if you are the hot hot thing.
But of course the great thing about prediction systems is they learn from new data.
