How Gopi got into investing

I was born and raised in Bangalore, and growing up, I was the kid who was always building stuff. The first startup I joined was in India: Cellworks was trying to build a sort of digital twin of living cells by modeling biological interactions and reactions within cells through differential equations. It was a fascinating problem, but it had its limitations as you expanded the breadth and complexity of the cellular pathways, so I went to the University of Cincinnati to do my PhD and try out newer modeling methodologies like agent-based modeling and network modeling. 

‍

After that program, I worked at IBM Watson Health as one of their first data scientists specializing in Real World Evidence and Watson for Genomics. Then, I met Collin West, who was a founding member trying to build a data-driven strategy at Correlation Ventures. They were pioneers in trying to do Moneyball for VCs. I found data-driven VC space an interesting new challenge that had a lot of similarities to my prior research work, so I joined Collin in building this out for the Kauffman Fellows Fund. During our time there, 50% of the companies we invested in together became unicorns. We knew something was working, so we thought: Let's build this a little bigger. Ensemble is the next iteration of our data-driven strategy. 

‍

Key opinions

Data has its limits… I spend my life building prioritization models, and our models are great at saying: Here’s a list of 50 compelling companies. What humans are great at is understanding the value of what’s being built at each of those companies. We use data for sourcing, screening.. anywhere we need data to help with prioritizations but when it comes to evaluating companies, the usefulness of the data has run its course. There will be strong disqualifiers that a model can never predict: Was the founder unprofessional? Is the customer reference subpar? No amount of data will prepare you for that. Investors make judgment calls based on hunches and trends and the perceptions they form of founders and their customers. Those are critical human evaluation points. They always will be for VC.

‍

…But data should be central. When we started Ensemble, I was very clear that data would be the core of our strategy. We made an early decision that we weren't going to be a software firm; we were going to be a data firm. We weren’t going to spend tons of resources on a fully-built front end. We would use low-code platforms. Right now, our team is uniquely a 50/50 split: We have as many data engineers as we have investors. There aren’t many firms that have that breakdown. But this was our bet, and it has paid off.

‍

Organizations aren’t single entity—this is Ensemble’s thesis. Our worldview is that it’s all about the team, market, and product at early stage startups. So we spend all of our time on the data team trying to understand those things. The goal is to provide the investors with a holistic, multi-faceted view of a company and its team before we even meet the founder. Investors generally can’t get the information they need to make an informed decision from a few calls with the company.

‍

Key outlooks and predictions

A source-of-truth platform will be table stakes for firms. The way people share information in VC is generally broken: Calls are made, texts are sent, someone tells you about a deal, and that source isn’t captured anywhere. Every time I talk to a VC firm, they say their biggest challenge is that not everyone in the firm sees the same thing. Everyone has their own CRMs, and the data is siloed in texts and inboxes. This undermines the objective of a data-driven system. Two years back, our KPI was that 90% of everything that happened around a deal would be captured in our platform. That was a huge forcing function. When all the data is brought into the same view and the firm has a single source of truth about what’s happening, they can more accurately evaluate a company. Firms will have to figure out that central platform if they haven’t yet. 

‍

Process will be just as important as data. You can have the best model out there, but if there’s no process in place for what happens after a list is shared, the data is meaningless. So the currency that Data Science has to have with the investment team is trust. Investors have been working in this space for decades, and if they're going to depend upon a model to make such high-impact decisions, you have to get their buy-in. Early on, we sat down with the investment team and explained every algorithm to them before we asked them to use it. I have enough Lucidcharts to overwhelm anyone because we described every piece. Firms are becoming more and more dependent on data. But process will have to accompany that data if they want to maintain an edge.

‍

Then there’s the investment process itself. There needs to be transparency and understanding between the investment team and the data team on which parts of the investment process are actually helped by data models. A lot of people are waiting for the day that AI will select unicorns the day they’re founded. The truth is that data is already supercharging investor capabilities; it’s just not in the parts of the investment process that you might expect. So we are very clear about allocating computer work to computers and human work to humans. Computers aren’t good at making the call to invest or not invest 

‍

“Two years back, our KPI was that 90% of everything that happened around a deal would be in our platform. When the firm has a single source of truth about what’s happening, they can more accurately evaluate a company.”

‍

Automation and AI will elevate firms. AI is already making a significant impact on knowledge work, and it will influence VC. There's so much unstructured data VCs consume that can be structured to organize what they're looking at and help them see things better. Internally, we’ve calculated how much improvement our models have seen with AI; and it’s only going to get better, cheaper, and faster. If you haven’t figured out how AI will influence your strategy, start spending some time there. For me, it's a productivity gain. It gives me the leverage I need to spend time where I feel good spending it. If I have five meetings a day, they should be the best meetings I could possibly have that day. 

‍

The advice I’d give my younger self

Honestly, I’m not sure how much I would tell him. There’s something about the process of failing and learning that’s so critical to this work. Data is by default an iterative process, a neverending exercise. I’ve made so many errors—not negotiating well with certain vendors, building tools too early when the data wasn’t right yet. But at the end of the day, error and iteration were required for us to get where we are now. For example, a lot of the projects we undertook were technically useless. But then we broke down the components of some of those projects to build some other tools that we’re using now. A lot of our current tools had their genesis in earlier failures. We’re notorious for killing projects. If we don’t see traction within a month, we’ll kill it. And we’re pretty ruthless about it: We only have three products at a time, and if we ever have one more than three, we’ll kill one of them. When I joined Collin seven years ago, my level of understanding of the data and the domain was so low. It’s only by getting your hands in the data, making mistakes, pulling projects apart and Frankensteining them together again, that your understanding of the domain becomes so rich. I’d want my younger self to experience that all over again.