1/ I hear confusion about difference between systematic VC and data driven VC.
a) Systematic VC will find/evaluate co's algorithmically. Decisions based on set of rules.
b) Data driven VC has typical investment committee making decisions informed (not instructed) by data
a) Systematic VC will find/evaluate co's algorithmically. Decisions based on set of rules.
b) Data driven VC has typical investment committee making decisions informed (not instructed) by data
2/ Because of the lack of data historically in VC decision making, 2b sounds novel. 2a I’ve talked about before.
3) ~25 venture firms right now are saying publicly they are leaning into data in “meaningful” ways -i.e. hiring inhouse engineers, data scientists, etc. Another 50-75 are doing it privately.
4/ Everyone in the ecosystem – the GPs, the LPs, the entrepreneurs- are loosey goosey about what it means to use data. A firm might hire 1-2 data analysts, call them data scientists and consider that in the same league as Renaissance Technologies. Nope.
5/ When evaluating the use of data, focus on the specific problems they are trying to solve with data and the resources they are putting behind those problems. In VC it effectively boils down to:
a) Sourcing
b) Winning the Co.
c) Evaluating the Co.
d) Helping co. post close
a) Sourcing
b) Winning the Co.
c) Evaluating the Co.
d) Helping co. post close
6/ Sourcing – Most VCs are scraping LinkedIn and Glassdoor to identify where engineers and data scientists are moving. I view that as a derivative. Whole lot of tech co’s have hired great engineers and it didn’t work out (survivorship bias makes this hard to see).
7/ Problem for the VC is that a) the business models in tech are varied, not uniform (thus hard to test or rely on the same factors), and b) there isn’t a lot of data publicly available on most companies.
8/ “What about credit card data?” First the data isn’t relevant for many tech categories. 2ndly where it is relevant, most sources of credit card data can give you the info only on a few thousand companies.
Thus- hard to use for sourcing. Also - quickly becoming commoditized
Thus- hard to use for sourcing. Also - quickly becoming commoditized
9/Winning the Co- In VC just because you are able to identify the right company doesn’t mean you’ll win it. This is a core issue I have with how much of an edge the tech can really give these VC firms.
10/ If some new VC firm invests millions to build a sophisticated data engine that gives them an Information Edge, they still have to the convince the co. to work with them. Wont that company still call the 5 best brand name VCs? I would……
11/ …..And I did. Several yrs ago I got a call from one of the early VCs to use data. They liked what they saw and could close in a few days. I thanked them for their call but still worked with USV. That firm couldn’t win the deal because USV brought much more value outside of $
12/ So what we’re beginning to see is VCs trying to win the deal with bespoke data projects to impress the entrepreneurs. I guess I don’t see this as particularly disruptive but it can be valuable to the entrepreneur.
13/ Evaluating the co- here is where the data is falling down but the LPs aren’t poking holes. First – these VC firms are almost universally still using human driven investment committees to make the decisions.
14/ How do you assess a self driving car company with data when there has literally never been a successful self driving car company? Read as ZERO TRAINING DATA. That is not a systematic fund.
15/ In the public markets there are discretionary managers who use data. And there are systematic managers who use data.
Key difference is systematic makes the decision algorithmically.
Same will be true in VC.
Key difference is systematic makes the decision algorithmically.
Same will be true in VC.
16/ So what do we have then? The same human heuristic decision making process, just with them spending a bit more on data? Well that doesn't sound like a fun party.
17/ Helping the co. post close – I heard a good quote today:
“[Top 3 venture firm] has quietly hired several data scientists. Basically all they are doing is one off projects for their portfolio companies.”
Hmmm.
“[Top 3 venture firm] has quietly hired several data scientists. Basically all they are doing is one off projects for their portfolio companies.”
Hmmm.
18/ I think data – whether used by a discretionary vc fund or a new systematic vc fund – can add value if used effectively. But what is key is that there is a clear articulation of what the problem to be solved is with the data.
19/ If you are evaluating the VC firm -as an LP, co-investor, entrepreneur, potential teammate- get specific about which problems they are solving with the data and how. And sometimes more importantly, which problems aren’t they solving with the data.
20/ Next step is to evaluate the resources behind their data engine. Think of this as a rough sanity check. Can two data scientists (who were data analysts a day ago) really build an evaluation machine?
21/ What if your VC firm has a few data scientists and no engineers? Ask where the data comes from. I’ve never met a good quant investor that wouldn’t say the data is much more important than the models.
22/ So how do you get great data without engineers? Buy it. (meaning everyone is likely buying same data) There isn’t a way to get non-commoditized scalable and reliable data over time without data engineers. At least I’ve never seen it.
23/ IMO data scientist are great at cleaning data for a specific use case, analyzing them and extracting information but most are not going to be experts in building reliable data and processing infrastructure that can ingest data over time.
24/ The ingestion and processing infrastructure is crucial in enabling data scientist to extract value from the data much quicker than trying to do it all themselves.
25/ Also insights cannot be a one time thing. Which means fundamentally you would need engineers to maintain things over time.
Your vc firm needs some engineers. Could be 3:1 ratio of engineer:ds for this problem; depends on a lot of inputs. But it isn't 0:1.
Your vc firm needs some engineers. Could be 3:1 ratio of engineer:ds for this problem; depends on a lot of inputs. But it isn't 0:1.
26/ Problem though- the VC for some reason only wants to hire data scientists? In my experience that’s because what they are really doing is acting as data analysts. They aren’t searching for factors that are predictive of success.
27/The market will evolve. I see VC and PE firms alike leaning meaningfully into data.
What is critical in your evaluation of the VC is digging deeper into 1) what problems they are solving with the data, and 2) what resources they are willing to put behind the problems.
What is critical in your evaluation of the VC is digging deeper into 1) what problems they are solving with the data, and 2) what resources they are willing to put behind the problems.
