HBS VCPE Conference Recap: Why Data Science will Improve, NOT Replace, Human Judgment

Speaking at Harvard Business School’s Annual PEVC Forum in February, a panel outlined how PE is leveraging data science

Speakers at Harvard Business School’s annual private equity conference discussed the rapid changes facing sponsors today, some cyclical in nature, some secular. On my panel, focusing on AI and Machine Learning, we highlighted the ways data science can help solve the industry’s greatest challenges, without replacing the very people who make PE tick.

By Sajjad Jaffer, Two Six Capital

To kick off Harvard Business School’s 25th Annual VCPE Conference, held in the last week of January, one of the first keynotes to speak at the event observed, “The [private equity] industry is changing really fast and heading to a place no one quite understands.”

In a sense, this sentiment perfectly summed up many of the discussions that followed throughout the day. The exponential growth of the asset class — whether measuring by committed capital, the number of firms competing for deals, or the diversity of strategies deployed — has, collectively, created new pressures that speak to the institutionalization of private equity, as well as the cyclical and secular challenges confronting GPs today. One speaker, for instance, remarked that entry valuations have settled into “grotesque” heights over the past several years, while another noted that it only takes two or, at most, three bad investments in a fund for the partners to “never raise capital again.”

It’s these pressures, however, that are driving the industry to rapidly evolve. But while disruption initially evokes more questions than answers, it can be comforting during these extended watershed moments to recall Blaise Pascal’s famous line, “It is not certain that everything is uncertain.”

Indeed, for those who took part in the Artificial Intelligence and Machine Learning panel at the HBS event, it is not lost on us that Pascal — the father of probability — would likely appreciate the role of data science to bring clarity and direction in navigating an ever-ambiguous PE landscape.

Focusing on the Statistics

In addition to myself, also taking part on the panel were Gradient Ventures Investor Tiffany Loer, Scale Venture Partners Principal Jeremy Kaufmann, OpenView Venture Partners’ Mackey Craven, Ubiquity Ventures Founder and Managing Partner Sunil Nagaraj, and Base10 Partners’ Rexhi Dollaku, a Principal at the firm. The panel was moderated by HBS Associate Professor of Business Administration Andy Wu.

While much of the discussion largely revolved around issues influencing VC investments in AI or ML startups, significant portions of the conversation touched on some of the key philosophical questions these technologies have stirred up. For instance, the panel — mirroring a larger debate across the industry — shared opposing views as it relates to the role of people when AI and ML capabilities hit critical mass. More pertinently, at least when it comes to spurring mainstream adoption in the near term, the discussion also reflected some of the challenges that can inadvertently stand in the way.

At Two Six Capital, we are not investing in data science businesses; we’re leveraging data science (through big data and analytics) to inform PE investments in more traditional industries, from retail and consumer companies to business services and technology companies. What was interesting though, and perhaps reflected the divergent perspectives on the panel, was how quickly the conversation descended into an “inside baseball” account around linear regression analysis and hyperparameter tuning.

It’s not that these more technical topics aren’t relevant to investing. For those targeting startups in the AI and Machine Learning segments, this granularity around process and modeling is indeed critically important. But in our experience, when trying to marry data science with business strategy, these more esoteric concepts tend to obscure the findings that matter most — the statistics that can tell a clear, unambiguous story around the unseen drivers influencing performance and returns on invested capital.

Over the last five years, for instance, we have worked on or co-invested in deals worth a combined $27 billion in enterprise value. What I think is more notable, though, is that through this work we have analyzed receipts totaling nearly $120 billion in combined customer transactions.

The most common revelation when first seeing data science in action — at least to the CEOs and general partners we work with — isn’t necessarily the breadth of the models we can apply, but rather the actionable insights around customer behaviors and product lifecycles that can then inform business strategy. Through this data, we can track all

athe different interactions between the business and its target audience, from a customer’s first purchase to all ensuing transactions; we can analyze increases or decreases in basket size; measure the time between purchases; assess how responsive certain segments are to discounting or marketing promotions; or gain understanding around lifetime value. The potential learnings are truly limitless and changing constantly. But when it supports a narrative as told through an ROI lens, these statistics can influence everything from resource allocation decisions to new business initiatives.

This is not to say that the model is not important. In fact, for those of us in data science, we live and breathe in the “modeling minutiae” in order to eliminate bias, interpret changes in the outputs, or continually improve performance. But to make the data actionable, decision makers want to know what the data ultimately says and the extent to which they can truly trust it.

Judgment, Informed by Data

This, of course, leads to the more existential question around the role of human intervention in this fast-approaching era of deep learning and neural-network technology. The potential of data science to recognize and discern trends that can’t be processed or observed by humans, alone, would seem to naturally lead to a point in which the robots take over. One fellow panelist, for instance, used an example in which he described how a model can be trained to observe a bottle tip over more two billion times, and if physics are incorporated into the underlying algorithm, “the need for judgment becomes more and more rare.”

What I believe this overlooks, however, is the inherent randomness of potential outcomes. Data, or more specifically big data, will certainly become an invaluable tool. In fact, it already is. We’re using it to help private equity firms perform due diligence and refine entry valuations, it is being applied to identify and pursue value-creation initiatives that offer most compelling ROIs, and investors have also begun to look at the data to inform exit timing to maximize returns and optimize the time value of capital.

All that being said, we continually see statistical models misbehave in real time, requiring continual calibration and refinement. And from the perspective of a fiduciary, as long as “black swan” events occur, oversight and judgment will remain fundamental to optimize risk-adjusted returns. (Anyone who thinks otherwise should revisit Nassim Nicholas Taleb’s seminal book, “Fooled by Randomness: The Hidden Role of Chance in Life and in the Markets.”)

Moreover, though, the learnings gained through data science are just that — insights that support and inform decisions. Human judgment, which differs from knowledge in a vacuum, is and will be required to prioritize what matters most to customers, investors, employees and any other constituencies being served. Human judgment — again different from knowledge — will ultimately serve as the bridge that connects the potential paths available with expected results to the specific and divergent outcomes most desired by the constituencies being served.

To go back to Pascal, he often cited that mankind generally suffers from two distinct excesses: “to exclude reason, and to live by nothing but reason.” This is why we believe that those investors who are able to “live in the grey” — understanding that nothing is black and white even with the benefit of hard data — will continue to outperform the competition. This isn’t to understate the role of data science, because, ultimately, it will become the foundation upon which reason can flourish and advance.

This article was originally published on LinkedIn in February.

About the Authors and Two Six Capital:
Sajjad Jaffer, along with Ian Picache, is a co-founder and managing partner at Two Six Capital. Sajjad, most recently worked with a private equity firm in Dubai and previously spent five years in a consulting capacity with Infosys. He currently serves on the Advisory Board for the Wharton Customer Analytics Initiative.

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store