Savvy sponsors are increasingly using data science to improve outcomes – from deal origination to due diligence and value creation.
Henrik Landgren was almost two years into developing EQT Ventures’ Motherbrain when he and his team realised the proprietary machine-learning platform had hit a tipping point in 2018.
Users had started sharing stories about how spooky it was that Motherbrain could predict interesting companies for them to meet. The platform finally had a good interface, which led to increased usage and, ultimately, to trust from users. Today, Motherbrain is an integral part of EQT’s deal-sourcing efforts.
EQT is one of a small number of firms leading the big data charge in PE. They primarily use it in three areas: deal origination, due diligence and value creation. Here is how firms on the cutting edge are using data science to try to gain a competitive advantage.
Deal origination
Sweden’s EQT and London-based Hg are among the first firms to embrace data science to find the best deals.Deal origination is largely a human endeavour. Personal networking and word-of-mouth are the dominant tools of the trade. In fact, only 3 percent of respondents in sister title Private Funds CFO’s Insights Survey 2021 said they use artificial intelligence for sourcing investment targets.
EQT, founded in 1994, began developing Motherbrain with the launch of EQT Ventures in 2016. The system currently monitors more than 10 million companies, pulling data from dozens of structured and unstructured sources to identify patterns that may be useful as EQT searches for target investments.
Over the years, Motherbrain has evolved from supporting only EQT Ventures to being used as the main dealmaking platform within all of EQT’s private capital strategies. To enable collaboration across EQT’s business lines, its relationship intelligence scoring highlights internal experts within EQT for any given company, says Elin Backlund, deal engine lead at EQT Ventures. “One of our product principles is transparency. We will only keep information siloed if we have to.”
As such, Motherbrain knows who in EQT has meetings lined up with potential targets and pings them a list of the company’s investors, its affiliate companies, as well as similar companies in the field.
Motherbrain’s expansive data infrastructure includes crowdsourced data, third-party data from data brokers and financial data scraped from the web on companies of interest to EQT, Backlund says. “The idea is that we want to model the world. We want to have all the companies that are out there in Motherbrain. That’s the only way we can do sourcing and identify emerging trends, rather than going always with the traditional work of being reactive to incoming leads.”
EQT Ventures does not use Motherbrain to source all of its deals, but the platform has been pivotal in investments in nine of more than 60 US and European companies. They include Peakon, a Danish HR start-up; Handshake, a career network for US college students; and WarDucks, a Dublin-based augmented reality gaming start-up.
EQT notched its first exit of a Motherbrain-inspired investment on 28 January when Workday agreed to buy Peakon for $700 million. EQT was the lead investor in the company’s Series A in 2017 and actively supported the next two funding rounds, which helped the start-up raise more than €50 million over two years. Peakon was a company that Landgren found himself using Motherbrain’s early algorithm.
Landgren, who built the analytics team at Spotify before joining EQT Ventures as an operating partner, says the firm learned some important lessons in the early days of Motherbrain. For example, technology needs to be extremely easy to use and truly add value for users to embrace it. The first two years of development for Motherbrain involved a lot of iterating to make sure it became part of the daily life of every person on the deal team. “When the platform was not as easy to use [in the beginning], users instead went to other platforms to manage their workflow,” Landgren says. “That meant we didn’t get the data needed to train the algorithms, and we also couldn’t effectively get the team to act on the companies suggested by Motherbrain.”
Today, the EQT Ventures team has a 98 percent engagement rating with Motherbrain, Landgren says.
“You log in, you have your individual interface, which looks like the best-of-breed productivity tool out there, and you see the companies you are tracking,” he says. “And using signals, the algorithm helps the user prioritise the companies that are most promising to look at.”
Like EQT, Hg is a big believer in big data. The tech specialist launched a data analytics team to support its portfolio team in 2016. To be clear, Hg’s analytics team does not drive the decision about which companies to back. It aggregates as many different signals (such as emails and internal communications) and scrapes public sources (such as press releases from company websites and filings from the UK’s Companies House and regulators) to accelerate the deal funnel. The tool sends research and updates to deal executives to help them prioritise the best possible investment opportunities.
“We have over 10,000 targets in the pipeline and we enrich that by using public and alternative data sources to try and flush out where the best investment opportunities are,” says Christopher Kindt, a director in the firm’s portfolio team who also leads Hg’s 20-member data team.
“One of our product principles is transparency. We will only keep information siloed if we have to” Elin Backlund, EQT Ventures
Due diligence
Due diligence can be lengthy, laborious and inefficient, which makes it ripe for improvements via data science. Making the process quicker – without sacrificing quality – can give an investor an edge.
It is unclear how many firms use data science in their due diligence. Most of the experts we spoke to say data science is not yet a standard in due diligence processes. But that is starting to change as more and more investment theses are running straight through the technology and data functions – and in these cases, it becomes a critical component of how to execute a company’s plans.
Neuberger Berman’s private equity group is among those looking to leverage big data to enhance due diligence. It collaborates with the firm’s seven-member data team, says David Stonberg, global co-head of private equity co-investments at the firm.
For example, the PE group and data team have collaborated to address due diligence questions such as: is a portfolio company’s growth sustainable? How strong is its brand? How effective is its marketing strategy? “The data helped us determine the actual level of brand awareness and the company’s competitive position and helped inform our investment decision,” Stonberg says.
At Hg, the data team spends about a third of its time on the pre-deal/due diligence phase, Kindt says. “That’s firstly about seeing whether we can already identify what the value potential of data is within an investment to help firm up the conviction we have of an investment.”
Secondly, and on applying data science to do the diligence, Hg gets its hands on specific operational financial datasets to give the deal teams deeper insight around customer behaviours to find out what is really going on in the company.
Rather than build an in-house data analytics team to handle due diligence, some PE firms outsource the effort to data science companies like Ekimetrics. The Paris-based firm, which is backed by Tikehau Capital and Bpifrance, worked on its first due diligence project with a private equity client about four years ago. That effort used new data to bring a fresh perspective on investment sectors such as e-commerce.
Fast forward to 2021 and both the client type and scope of work have evolved, says Matt Andrew, a partner at the company. “We’re seeing more mid-market players looking to this type of expertise as a new lever for driving growth and value creation.”
Today, Ekimetrics speaks with senior management teams to find out how they think about their data roadmap and what they think about integrating analytics into business processes. Sometimes it is asked to judge the maturity of a company’s data assets.
“I think there is this greater appreciation that part of the value of a business can come from what you can do with the data,” says Andrew. “How you can monetise it and how you can create value in revenue generation or cost reduction through the deployment of data assets in the business.”
Value creation
Data science and AI generate the most value for private equity firms in performance analysis. Findings from Private Funds CFO’s Insights Survey 2021 reveal more than 40 percent of respondents indicated “medium” to “high” on the effectiveness of using machine learning to measure and analyse investment performance.
There are many examples of firms using data to boost their portfolio companies.
TA Associates built a prediction algorithm that looked at the behaviour of a B2B tech company’s clients over three years. It identified which clients were most likely to grow their accounts and shared the information with the company’s business development team. The tech company determined that the predictions were accurate in 96 percent of cases, says Catherine Cutts, vice-president, strategic resource group and head of data science at TA.
Hg’s investment with UK software business Access Group is another example of how a firm used data and AI to grow a portfolio company’s sales. Hg’s data team built a customer warehouse in an Amazon Web Services cloud infrastructure, then fed data from the warehouse into a machine-learning model that scored all cross-sell opportunities.
The model improved cross-selling output by more than 60 percent, says Kindt. He adds that Hg’s narrow investment focus – software and services – makes it easier to repeat this data-driven playbook than it would for a generalist private equity firm.
In some cases, the insights from data contradict expert opinion. Sajjad Jaffer, founder of data science firm Two Six Capital which was recently acquired by business and tech consulting firm West Monroe, recalls an example. A global software company’s private equity owner hired a consultant to determine if the company should expand in the US. The consultant deemed it a bad market due to oversaturation, brand recognition surveys and traditional methods of research. However, Two Six found the opposite was true.
“From this research, which looked very believable, they said the market would only be expected to grow 7-8 percent,” says Jaffer. “However, we looked at the client’s specific data and spotted a trend that their US customers were accelerating adoption of their product, and if this acceleration rate held it would be a very lucrative market for the company.”
Faced with the two opposing views, “the private equity firm ultimately took our advice and their US market business grew 100 percent year-on-year”, Jaffer says.
Data science in effect provides an additional lever to improve EBITDA and margins in a company, says Cutts of TA. This de-risking is done in two ways: first, by “squeezing any extra juice” and second, through scenario planning and forecasting.
“If you had to do lead scoring or marketing optimisation, that might improve things by 20 percent, and you might want to underwrite that improvement or keep that improvement as sort of a de-risking strategy,” she says.
On forecasting, Cutts says simulation techniques can model thousands or even millions of different situations to better understand risks. “An example of this might be forecasting demand growth in a niche market where there are no official forecasts,” she says. “A data science forecast might combine lots of different official forecasts – for example, interest rates, gross domestic product and wage growth – using their historical correlations with the niche growth rate to build a forecast that can take into account some very complex dynamics.”
Sounds great, but can I afford it?
Firms are increasingly leveraging the power of data science, but how they approach it differs based on size. Larger firms usually build the expertise internally, while mid-market players typically outsource it.
The Private Funds CFO survey shows the adoption of artificial intelligence, robotic process automation and machine learning is in single digits. For most respondents, the convergence of AI and private equity is sometime in the future. More than two-thirds of respondents expect AI to have the greatest impact in five years, and almost a third say it is a decade away.
A key reason for the slow adoption is building an in-house data science team is costly. Firms we spoke to would not disclose the exact cost, but all say it can be a large, ongoing expense, largely because the required talent is hard to come by.
Henrik Landgren, a partner at EQT Ventures, declined to say how much the firm spent to develop and scale Motherbrain, its machine-learning platform, but he notes that being part of a more than €50 billion asset manager has been an advantage. The Motherbrain team started with three data scientists and has grown to 20, including full-stack developers, system engineers and product designers.
Hg and TA Associates both work with contract data scientists and developers in their projects. Hg has a team of 10 data leads in London and teams in Romania and India to provide scale. At any one point in time, Hg’s data team employs another 20 or so contract data specialists and technical experts. For its part, TA has two in-house data scientists and brings in more experts to flex the team and bring the required skillset on a project basis.
The real barrier to entry is that data scientists are “pretty high in demand and pretty demanding as well with the jobs they are doing and the kind of opportunities they want”, says Matt Andrew, a partner at data science company Ekimetrics.
“The bigger risk is for a PE firm to try and build an internal team and start to have knowledge leak,” Andrew says. “You need to have a certain mass of opportunities and projects to keep your data science team engaged. We have seen big corporates build data science teams and just realise they eventually have to put three data scientists on a single project. Because of the churn you get, it’s not easy to scale from those working functions.”
Projects which cover a data assessment, deployment of a data roadmap and a team of up five data scientists can cost up to £300,000 ($415,000; €340,000), with the cost varying based on the scale of the ambition and timeframe, according to Ekimetrics. Multiply that cost by, say, seven companies in a mid-market firm’s portfolio and you can see why smaller firms experience sticker shock.
The human touch
It is important to understand that while data science can improve outcomes, it requires talented people to execute plans and make tough decisions.
“At least half the challenge is actually change management – changing the internal processes and operations within the business,” says Hg’s Kindt.
“The data science bit is in a way more linear, easier and predictable. How you then change behaviours within the company, that is tough and that takes effort. And that’s where having someone in-house to oversee and orchestrate and bring people along with you on that journey plays an important role.”
At the heart of the investment process is the deal executive. The data science team’s role is to serve up better insights for decision-making, not build an all-encompassing model that assesses or understands the overall investment risk, Kindt notes.
Similarly, in deal sourcing and diligence, there is only so far a data-driven platform can go. For AI to derive material insight, it needs human intervention.
Says TA’s Cutts: “Invest in people, not companies, as the old saying goes.”
How has the covid-19 pandemic affected your work?
Christopher Kindt, director in the Hg portfolio team and data team lead:
The pandemic has made management teams much more open to experimenting with new ways of working and seeing whether data science and AI can help improve a certain process. Previously it was much more common to encounter reluctance to review and challenge a set way of working.
If anything, there hasn’t been a better time to start introducing, experimenting and trialling all sorts of AI projects in companies because many things have already been forced to change. Many companies are already in the process of going about their business in new ways.
Matthew Kaplan, head of Almanac Realty Investors:
The Almanac team has worked closely with the data science team since we joined Neuberger Berman early last year. At the beginning of the covid-19 pandemic, Almanac engaged the data science team to help develop a better understanding of consumer behaviour at one of its underlying portfolio companies focused on acquiring and repositioning retail centres across the US.
The data science team used credit card data and geolocation/mobility data to measure trends at the underlying retail properties of Almanac’s portfolio company, including weekly sales, customer visits, spending patterns, customer profiles, etc. This provided the Almanac team and its portfolio company with additional insight on consumer activity at the underlying properties and how it changed over time as the pandemic progressed.
Ian Picache, senior director, West Monroe:
The value creation and diligence work we are doing has been focused on identifying how consumer behaviour has changed post-covid-19 and then how to capitalise on this change. An example of this was when we were working with a consumer technology business that saw increased growth due to the pandemic. The technology company had a freemium-oriented structure and through our research we found there was a great number of non-paying customers that should be paying due to how much they were using the platform.
We found if these users became paying customers, revenues would increase by 40 percent. To monetise this, we set-up usage blockers leading to a paywall. Understanding these important consumer behaviours post-pandemic is paramount. This was for a value-creation opportunity, but then we used the same process and methodology for a similar business during diligence for the private equity acquirer.
Source: Private Equity International