PE Quant Fund

Primary Fund selection, co-investment decisions and secondary stake investing are the “bread and butter” of PE asset management and the basis for billions of dollars of advisory- or management fees charged every year.

Can you imagine what would happen to this industry if a simple set of algorithms running on a standard computer would be able to consistently take better decisions in terms of Primary Fund due diligence, co-investment opportunity assessment and secondary fund stake pricing than the existing teams of skilled and experienced investment professionals?

It is natural to believe this will never happen.

Isn’t PE a people business after all? And isn’t PE data too scarce, too spotty and too unstructured for algorithmic decision making to work?

It may be time to reconsider this belief.

As an academic I worked with PE data for over 25 years. I have built (and published) the first prototype of a simple model to predict future returns of a primary buyout fund back in 2007. And I worked extensively since then to continuously experiment with additional variables and better methods.

The accuracy of the corresponding models improved steadily over the years, but things started to become fascinating when I was able to integrate advanced meta machine learning techniques that were designed to deal with the scarce, spotty and unstructured type of data that characterizes private equity.

That was back in 2018.

Seven years later, the results of thousands of carefully designed and “intellectually honest” back tests clearly indicate that PE asset management will soon be disrupted by algorithmic decision-making processes:

Had I launched a fund of funds in 2019 based on an algorithm trained with data up until 2018, it would be today a top quartile performer with an 21% IRR and 1.4x TVPI

Had I used an algorithm built with data from pre 2017 to filter co-invest opportunities from 2017 to 2020, the best 25% of these opportunities would as of today outperform the total sample of opportunities by 14% IRR and 0,3x TVPI

Had I launched a secondary PE fund in 2019, running on an algorithm trained based on data up to 2018, it would be today a top quartile performer with an 24% IRR and 2.04x TVPI.

Recently a friend asked me: “So what are you waiting for? Go ahead and find ways to apply your algorithms. This will start a process that will disrupt the status quo and improve the efficiency of the asset class as it will channel more capital into the best deals and the best funds to the benefit of investors.”

If you are curious to learn more, please check out my recent plenary keynote at the Super Return International Conference in Berlin: