The data advantage in secondaries (Part 2 of 2): Picking the best assets

We pick up our conversation with Wilfred Small, Co-Head of US Secondaries & Primaries, Senior Managing Director & member of the ASF Management Committee, and Arnaud Mercier, Managing Director, about how data science and AI support Ardian’s positioning on the secondary market.

Read Part 1 here

QUESTION 1.

How does Ardian’s Secondaries & Primaries team work with data scientists?

ARNAUD MERCIER - In our London office we have currently three data scientists physically located on the same floor than our Front Office team, in order to have them as integrated as possible with the investment team. They have two main missions. First, they clean, organize and enrich our data with web information, external databases or documents from GPs. That's 30% of their time. The other 70% is spent on analysis (ad hoc or recurrent), data extraction, as well as developing new AI tools to support the investment team. Ultimately, the success of implementing data science and AI within our expertise comes from identifying concrete use cases that can be either automated or optimized to increase the team’s knowledge, productivity and reallocate their time on more value-added tasks.  

In an industry where information is mostly private and covers thousands of companies, our data scientists’ work is critical. They are able to save weeks’ of work for analysts who previously had this workload on their shoulders, including recurring market analysis such as looking at the metrics of a specific sector, analyzing valuation evolution, EBITDA and sales trends, growth forecasts, etc. These analyses always generate a lot of insights on what is happening in the industry or our portfolio, and whether we share them with LPs or internally, they usually generate further questions or new ideas. Having people dedicated to the task is instrumental to come back with answers quickly.

For new or more complex analysis, our data scientists are working closely with our investment team. As an example, we explained that part of our strategy is to be highly selective about the assets we buy. For our June 2025 AGM, we wanted to look at how the performance of the funds we cherry-picked compared with those we didn't buy. And so we carried out a backtesting analysis within our most recent Secondary Buyout platform, ASF IX. We were pleased to see our selection was paying off: we bought portfolios at lower EBITDA multiples, with our selected assets trading -0.6x times below those we did not buy. But our selection had higher EBITDA growth and much lower listed exposure. As a result, our cherry-picked funds outperformed by 21.1% compared to the excluded funds, which shows the power of our database.  

(1) Performance of cherry-picked funds since entry record date across ASF IX transactions, including: Camellia II, August, Parsnip, Rembrandt, Universal, Crabapple, Cabana. Gross performance unlevered.
(2) EBITDA growth based on historical CAGR at EBITDA level from ASF entry to latest RD available.

QUESTION 2.

How is AI changing the way the investment team works?

ARNAUD MERCIER - I would say we are still at the beginning of the journey, but things have been already moving very fast over the past 12 months. One major step was achieved at Ardian when we launched our internal proprietary large language model Gaia end of 2024, allowing users to use generative AI to question web contents or proprietary documents which have been uploaded, whether it’s from a deal’s data room or investment memos. At first, most of our AI initiatives were focused on increasing the investment team’s productivity and automating recurrent and time-consuming tasks. As such we have developed two AI-powered tools: one helps in company analysis by finding the best comparables within both public and S&P database universes, while another helps to generate some draft wording for our company sheets based on GPs’ reports or presentation. Last but not least, we have automated our funds’ benchmarking tool, aggregating data from both internal and external sources, which is a key component of our analysis for primary candidates. With these quick wins implemented, we have now been focusing on using AI to enhance our level of due diligences, for instance leveraging on Gaia to support investment teams with a list of questions to ask when working on a targeted asset. 

WILFRED SMALL - We see AI as enabling people and optimizing their time spent. There are certain analyses that only AI can do because of the sheer quantum of data being processed that wouldn't be feasible for an individual to spend time on. But it provides an additional lens through which to make decisions, as opposed to making the decisions on our behalf. 

QUESTION 3.

How do you see things going forward?

WILFRED SMALL - I think the tools that we're developing are going to continue to become more sophisticated and more helpful in the investment process. And importantly, the quality of the underlying data is what is going to enable those tools. The quality of the data is a natural barrier to entry to our positioning on the secondary market. It takes a lot of time to assemble a portfolio, a team and relationships at that scale, and we've been doing it for 26 years. New entrants tend to lack the quality of relationships and therefore the quality of data. That's why we are the largest player in LP-led secondaries – because there are large economies of scale and significant barriers to entry. 

ARNAUD MERCIER - We are not lacking ideas in our data science and AI roadmap but need to prioritize based on what we think is critical for the business, The 3 main key focus at the moment are the development of (i)  a “Data Case” which would complement our team’s bottom-up investment case when pricing secondary transactions, (ii) a secondary deal snapshot overview which would quickly rate the attractiveness of a deal for us based on its main underlying metrics to quickly decide if we should spend time on it and (iii) continue to feed Gaia with even more KPIs and documents from various sources to become our “360 intelligence agent” that anyone in the S&P team could contact for any questions related to our activity and get an instant answer.