crunchscoring: predicting future startup winners with machine learning and crunchbase data
DESCRIPTION
At #oohack Hackathon in Helsinki in Oct 2012, Exacaster team decided to verify a few startup myths by employing machine learning techniques to analyze CrunchBase data. Some of the questions they addressed: - Is it true that only 2% of startups ever exit? - Are we getting better at doing startups over time? - Is it possible to improve your chances of picking the winners by employing publicly available data?TRANSCRIPT
What’s your CrunchScore? Tes3ng Startup Myths with CrunchBase Data
#oohack Helsinki Data Hackathon, October 16, 2012
Exacaster Team Sarunas / Egidijus / Rokas / Andrius / Vidmantas / Justas
[email protected] | www.exacaster.com
Only 2% Of Startups Did Ever Exit
founded 75%
raised funding 23%
acquired or IPOed 2%
Are We GeJng BeKer at Doing Startups Over Time?
Ignore – not enough Mme to exit, yet
Year company formed
Spot The Flood … And The Lack of Exits
Math to the rescue!
Our CrunchScoring Algorithm + Startup CrunchBase profile = Double The Chance of Picking the Right Startup
0%
1%
2%
3%
4%
5%
6%
7%
8%
9%
Top 500 scorers sold Remaining companies sold
8% of companies later exi3ng within the cohort with a Top
Predicted CrunchScore
4% of companies later exi3ng within the cohort with a Low
Predicted CrunchScore
Brought to you in 10 hours by:
#oohack Helsinki Data Hackathon, October 16, 2012
exacaster.com