When I first met Mattias Ljungman, he told me that he wanted to build the most quantitative venture capital firm in Europe. Equipped with data, algorithms, and custom-made software, Moonfire would set the standard for a new paradigm in sourcing, evaluating, and managing investments.
The vision was to establish a new form of venture capital where a team powered by software enhances its ability to execute with better knowledge and greater velocity than anyone else in the industry.
In a nutshell, that’s why I am here. I recently joined Mattias and the team (please see my bio here) and I want to dive in a bit deeper on what Mattias’ vision means.
VC firms talk about being data-driven all the time. What they mean is they use data to inform and accelerate various decision-making processes in venture capital. Typically, in venture capital, data is used to accomplish one of the following sets of objectives:
- Sourcing: Finding new investment prospects in the form of both founders and companies.
- Evaluation: Helping to determine whether or not new investment prospects would be worthwhile investments.
- Enrichment: Automating the most mechanical aspects of the research processes by methodological enrichment and annotation of investment opportunities with competitive market data.
- Investment: Allowing investors to more effectively accomplish their objectives by making decisions based on accrued knowledge of the performance of an individual, company, the performance of the portfolio, and the performance of the wider market.
- Support: Helping portfolio companies accomplish their objectives based on the acquisition and leverage of unique market data.
- Reporting: Ensuring that LPs are informed by leveraging an enhanced understanding of the performance of the firm and the portfolio.
These are really important objectives. And at Moonfire they are our objectives too. But we want to go further. We want to do more. We want to be a truly data-driven venture capital firm.
In even the most advanced data-driven venture capital firms using machine learning to evaluate companies, I’ve noticed an unfortunate reality. In almost all cases, these firms likely have a single statistical model which they’ve trained to classify potential investment opportunities and output a numeric score based on the prior success of existing companies with similar attributes. This value factors into their decision-making process but their operations are overall extremely traditional, just with a classical machine learning model in the loop.
At Moonfire, we’re building in effect an ML-driven fintech company which specializes in venture capital. We aim to use data, software, and machine learning to inform every level of the stack.
This means that we’re not just creating a single model and integrating it into a traditional pipeline. We’re building out operations from the ground up and approaching the problem like a software development project.
Our sourcing, screening, and evaluation process is a production-grade data pipeline. Throughout this pipeline, we make a series of decisions. Whenever we have an opportunity to use a learned model to facilitate better decision-making, we do so. When we have a series of companies or founders to filter through, we reason about the problem as if we are building a modern search and recommendation system.
We are writing our own software at Moonfire but we are also ensuring that we use venture-tech tools where appropriate. We aim to leverage the entire venture-tech ecosystem, creating a cohesive experience with our own software, and enhancing every tool that we use with our proprietary data solutions.
There will be more details soon. But for now, I’m exhilarated by what we’re building at Moonfire and I can’t wait to share more with you soon.