The book and movie “Moneyball” portrays how Billy Beane, the general manager of the Oakland A’s baseball team, used a data-first approach to build a team for the 2002 season at a third of the cost of other Major League outfits.
New York-based Ducera Partners is applying a similar logic to venture capital and investment banking.
The firm’s chief technology officer, Thomas Thurston, told DigFin, “We can’t disrupt Goldman Sachs by just being a mini version of them.”
But putting artificial intelligence at the heart of the firm’s process can increase productivity per person and allow the firm to do new things.
Deep roots
Ducera launched in 2015, but its leadership team have been working together since the late 1990s at WR Hambrecht & Co., an investment bank founded by Bill Hambrecht, formerly known as Hambrecht & Quist.
H&Q was a San Francisco investment bank founded in 1968 that specialized in technology companies; it underwrote IPOs for the likes of Apple Computer, Genentech and Adobe Systems.
Thurston, a data scientist, has been practicing at WB Hambrecht and now Ducera since 2008; prior to that he worked at Intel.
The proprietary model
Ducera practices venture investing, M&A and investment banking. Its business resembles traditional players but it relies on a proprietary algorithmic modeling tool called Quannix in order to estimate the valuation of a private company.
It doesn’t look at financial statements, revenues, or last-round valuations. Instead it samples more than 1,000 public companies and runs a thousand different learning models against them, to come up with a means of predicting a private company’s worth and growth trajectory.
This is based on public information available online or commercial data feeds, as well as the firm’s proprietary data mining and a database of large companies’ partnerships and research projects.
Thurston says Ducera doesn’t rely on sentiment signals like social-media likes or tweets. He likens establishing estimates on private companies to guessing the size and direction of a rock thrown into a pond by measuring the ripples it creates, rather than trying to observe the rock itself. “Every private company and its people leave a data trail,” he said.
The firm then graphs that market-cap estimate and direction onto a database of related or adjacent companies worldwide, so it can generate ideas around mergers and acquisitions or how the company can practice financial engineering to get a result.
Leaner origination
This replaces having teams of expensive bankers flying around the world to originate deal ideas. It also makes for a powerful pitch to clients.
“We’re not presenting them with a PowerPoint presentation decorated with six company logos. We’re showing them six thousand companies in a dynamic ecosystem,” Thurston said.
Since its founding in 2015, Ducera has advised on $850 billion of transactions, and is now doing about $100 billion of deals per year. (In 2023, Goldman was the biggest M&A bank, advising on 235 deals worth $671 billion.)
AI-based venture investing
Although the M&A business is the largest part of Ducera’s book, it is also making inroads in venture capital. Its first VC fund is now entering its ninth year, but the firm has many such portfolios.
Two things make Ducera different from most VCs. First is its AI-based approach to finding startups. Second is its business model, with every firm a co-investment in partnership with a large strategic corporate. The one listco that is public about this arrangement is Corteva Agriscience, a $41 billion agriculture company.
Thurston explains Ducera brings both a VC approach, seeking high multiples, and a corporate VC approach, looking for a strategic fit with its client LP.
Its algorithms allow it to cast a wide net for startup investments; the venture fund invests in seed to Series A opportunities. Thurston declined to say how many venture assets the firm manages as a GP.
Having the financial goal of a VC is just as important as a strategic relationship for startups. Thurston says CVCs always want to build an M&A pipeline, but Ducera’s research shows fewer than 1 percent of CVC portfolio companies get acquired by the corporate.
“M&A is not an output of corporate VC,” Thurston said. “It just doesn’t happen, and we tell this to the portfolio companies up front. Data isn’t just about doing deals, but understanding the drivers of a company’s innovation and growth. If you understand that, you can help engineer the outcome you want. It’s more like actuarial science than vague, hand-wavy stuff.”
The future of private investing?
This harks back to the Moneyball comparison. In the book and the film, Beane had to confront scouts and managers who relied on experience and intuition when building a roster or calling a play.
Thurston says there are only a handful of VCs using AI to this extent. “VCs like to think of themselves as artists,” he said. “Bill Hambrecht had no ego, and he was willing to use models that did a better job than he could. But for every Bill, there’s another nine investors who only feel existential anxiety when they look at what AI can do.”
He believes the industry is beginning to shift, with new tools for generative AI making it much easier for people to use these tools. “Private equity and venture capital investors will realize they can benefit from AI and data,” Thurston said. “It’ll happen.” But for now, he reckons there is about a dozen of serious data-science led firms, out of about 10,000 VC firms worldwide. “We’re a blip,” he said.
Once Ducera calculates valuations of private companies, be they startups or large scale, its work becomes more conventional. In VC, if it finds a company with interesting metrics, Thurston or a colleague sends the founder an email, cold. Usually they respond because of the strategic corporate that Ducera is partnered with. The traditional due diligence and company meetings then follow.
But Thurston says Ducera has a distribution model that varies from the traditional VC’s. To continue the Billy Beane example, most VCs expect most of their portfolio companies to strike out, a few to hit singles or doubles, and one smash a grand slam. Ducera’s algo-led approach means fewer misses, according to Thurston.
He notes that big investment banks like Goldman have the budgets and the talent to be big players in AI – but they are more focused on public markets. Hedge funds such as Renaissance Technologies and DE Shaw have been using algorithmic-led trading very successfully for decades – RenTech was founded in 1982.
But Thurston says most companies around the world are private. Some industries are populated almost entirely by private companies. That goes for big companies too, not just startups: Thurston says four out of five companies with revenues greater than $100 million are private.
So this is where Ducera is playing, using its algos to make guestimates to understand these markets, despite – or thanks to – the lack of public information.
