Correlation Ventures uses predictive analytics and data from 60,000 financings since 1987 to evaluate investment opportunities more quickly than traditional diligence processes. They consider about 15% of the 1,650 deals they screen each year. Their process involves filtering deals, applying a 30-factor model to score opportunities on a 100-point scale, and conducting diligence on deals scoring 70 or higher. They have made 48 investments so far at a pace of about 3 per month, with 40% in life sciences. Recent exits and improved industry returns are positive signs for venture capital.
BoxWorks SuperSession: “The Future of Healthcare and the Cloud”
Predictive Analytics in Venture Investing
1. Predictive Analytics in Venture Investing
Hosted by Correlation Ventures
4-17-2013
Industry Returns
25 year VC returns: 16.5% mean, 3% median
Are there “just 15 deals” that matter each year? The data shows that there are 400-600 5x
returns per vintage (3 yrs) so more like 200/year
Low (45%) repeat rate of being top quartile in fund returns from one fund to the next
35% of capital deployed created 45% of winners (>5X returns)
One third of rounds that were undersubscribed led to one third of the winners
Winners broadly dispersed by sector stage and manager
Geographic proximity to portfolio companies does not improve returns, actually the inverse was
true (surprising! – or maybe not…)
History of Correlation
Inventor of Falcon (fraud detection backed by Greylock and battery; acquired by Fair Isaac)
recruited to run analytics, got together with longtime life sciences and tech investors (David
Coats, Trevor Kienzle)
30 VCs personally invested including former head of NVCA
Recruited Steve Caplan and Matt Rosencroft as advisers
30 funds have provided peer data for exchange for fundraising analysis data
o Want to be the industry flag bearers to help people raise capital
o New research emails come out once per quarter.
Have records for 60,000 financings since '87 (90% of financings and 2/3 of returns)
Investment Process/Parameters
Decisions in two weeks or less
Don't repeat diligence that the lead has done; overlay a fast analytical approach - no fly outs
reqd from mgmt team or customer calls
Need five documents - 5-10 min process from entrepreneurs
Vote with lead; tells team what the reserve is
Can go up to $2M on first check, $4M total (can go as small as $50k)
o Pay-to-plays not an issue, plans reserves appropriately to follow on but would get
converted to common if no other choice (doesn’t expect that to happen in too many
deals and does reserve a slush fund to prevent good deals from running away)
o Thinks that there is no difference between how much they own and how those
investments perform
o Making smaller investments decreases portfolio risk by enabling diversification
opportunities
Data Processing
Filtered out: insider rounds, strategic led, non-VC deal, inexperienced VC led (lower quartile by
activity over five year period)
Factor model: financing, syndicate, lead (firm and partner)
Thirty factor empirical weighting; 100 PT scale
2. o Need a score of 70 to consider; about 15% of these make it
o Then call with lead VC and team
o Do background checks and legal work at own expense
Looking to ensure unusual factors aren't popping up (ie: criminal; lead VC is
mgmt and investor – surprised ALDEA made it through the process!)
48 investments as of Friday (pace about 3 per month); 40% of dollars are in life
sciences (recent exit: RQx w/ Avalon Ventures, sold to Genentech)
Can follow performance of “anti-portfolio” – the deals they screen but pass on
(Correlation’s LP’s will be interested in this data)
Good News
Number of financings has increased (near peak since 2001 bubble)
Can consider 1650 deals each year - wants to see as many as possible
Active VC firms bottomed out in 2010
Capital raised by US VC funds: $21B in 2012
Returns have improved dramatically 10% (those at 90th percentile) did 5.5x or better last year
o Mean was 3.1x last year, up from 0.4x post-bubble
o Driven by lower pre money valuations, higher exit values, more capital efficiency
Pre money valuations: 25% reduction since 2004
Exit values: 90th percentile exit in 2001 was $50M, in 2021 it was $300M
Capital efficiency: $21M in 2004; $18M today (explanation for tech co’s: push
off server expenses to amazon)