Predictions for machine learning in Private Equity

18/08/2021
Read Time: 4 Min

With machine learning driving growth across both private equity portfolio companies and private equity houses themselves, we asked Calum Mackenzie from AI and ML specialists Faculty to make his top predictions for the future of the technology in the PE industry. Faculty has advised many investors (including ECI Partners) on private equity investment in machine learning and portfolio companies on their AI implementation, so is well placed to predict what is likely to play out over the next year. 

1. Investors will continue to invest in ML capability (internally and externally)


Most PE houses now view ML as an inevitable force in their industry. We typically see funds adopting one of the following strategies: 

  1. Be a fast follower: wait for competitors to establish the optimal approach
  2. Buy in expertise: engage a company like Faculty to identify ML opportunities and guide their implementation
  3. Bring in a senior hire: bring someone on board to oversee strategy/implementation
  4. Build capabilities in-house: Build their own data science team and explore ML independently 

The easiest place for investors to start is often with ML due diligence and/or value creation within their own portfolio, so this is still where I’d expect the most investment in the short term. Origination tools like ECI Partners’ Amplifind™ have their place, but can require larger budgets, a willingness to invest over 12+ months, the right data and an R&D mentality to be successful. 

2. Investors will invest more in their portfolio’s ML strategy and roadmap

In June last year, The Economist ran a series of articles in their Technology Quarterly outlining that, although AI is expected to add $16trn to the global economy by 2030, many sectors and organisations have really struggled to get to grips with ML.

To avoid this trap, investors need to work with portfolio management teams to help them become more data savvy and outline a vision for advanced analytics and ML. Typically this means a progression from traditional spreadsheets/BI and rules-based software (executes rules deterministically) to the deployment of learning algorithms (takes data about the world around it to draw inferences probabilistically). Right now, the vast majority of portfolio companies are using the first two methods. Investors and portfolio companies that can accelerate deployed ML at scale will see this translate into portfolio company growth and in turn investment multiples.

3. Increased use of specialist Machine Learning Due Diligence (MLDD)


Historically ‘data and analytics’ due diligence has been covered by the commercial due diligence workstream. We are seeing more investors commission specialist MLDD where data science is material to their investment decision. For example: 

  • The company’s value proposition is predictive software
  • The company is in a sector where ML is having (or could have) a big impact
  • Large amounts of unstructured data need to be ingested to inform the decision

This is something we saw more of in 2020. For example, we advised ECI Partners on their acquisition of predictive analytics software provider, Mobysoft.

4. Investors will play a more active role in ML use case identification in the portfolio

Regardless of how active a given investor is with their portfolio, there is a growing feeling that ML use case identification is either too slow or not happening at all. ML will not be relevant for all portfolio companies, but there is great opportunity for investors to offer portfolio companies more support in understanding where it will drive value.   

For example, in summer 2020 we partnered with a B2C e-commerce fashion retailer to move away from rules-based retention marketing using a propensity model approach – a £90m annual revenue opportunity. In terms of use case identification and model build, this was a collaborative project working alongside the investor.

5. Investors will begin to understand the importance of ML-Operations

ML-Operations, or ML-Ops, is the ability to maintain machine learning models reliably and efficiently; it’s generally not that well understood by investors and portfolio companies. Implementing ML successfully means models are being used to make or inform actual business decisions or actions. Getting this right confronts portfolio companies with challenges they have never faced before. What happens when predictive accuracy decreases? What alerts and monitoring do I need in place? How will I audit my models? As investment in the ML space increases, these questions will become increasingly prevalent. 

6. Investors will start asking the portfolio questions about AI safety

Most people have heard the Amazon case study on discrimination in recruitment. But I would hazard a guess that you haven’t heard of B2C2 Ltd v Quoine Pte Ltd, which required the defendant company to determine knowledge and intention in its AI decisions. 

As more companies adopt AI, issues of AI safety are increasingly entering the public consciousness. Today, organisations have a much clearer idea of the potential risks of implementing AI incorrectly – from biased algorithms making discriminatory recommendations, to models that degrade and become less accurate over time. 

Managing the potential risks of AI deployments will become increasingly top of mind for investors and their portfolio companies as more of these examples come to light.

7. Portfolio companies need to decide where they’re sourcing their ML capability

Portfolio companies will increasingly recognise that their people are making predictions on a daily basis, typically without the right data. Companies need to establish what the most important predictions (churn, product demand, pricing etc) are for their business/sector, and then decide whether they want to build in-house capability or outsource it. 

The sooner companies make that decision the better, as the economic value of these predictions will reduce over time as ML becomes more sophisticated and widespread. You can either pay-as-you-go with a tech company – what we call AI-as-a-Service (AIaaS) – or take the DIY approach to ML. 

Both have merits, although it is worth making sure that you kick-off with a strategy project to work through the implications of either route in detail. Too often companies attempt a DIY approach and make painfully slow progress, but if they get it right from the get-go, they can gain a competitive advantage and deliver value at pace. 

Calum Mackenzie

About the author

Calum Mackenzie

Calum has 12 years of consulting and industry experience. Before joining Faculty, he was a Director in the consulting division of Iris Worldwide. In this role he ran the European Data Science and Consulting relationship for Starbucks. He also managed Data Science projects for Samsung, Shell, Value Retail, IHG Hotels and Genting Casino. Prior to Iris, he worked in various strategy and commercial roles at Amazon, BT / BT Sport and Capgemini Consulting. Calum has a Bsc Hons in Business Administration from the University of Bath.

ECI announce close of latest fund

Find out more