Managers may make better informed decisions with the help of machine learning algorithms, which can reveal hidden patterns in massive data.

Most merger, acquisition, partnership, and diversification decisions made by corporations are undertaken in the face of substantial uncertainty and intimidating complexity. As a result, it’s crucial to have a solid plan in mind in making such judgments, so as to navigate the inherent complexities and glean as much information as possible from the signals generated by the plan’s execution. Many people have come to realize that disappointing or even disastrous outcomes are more likely to occur when decisions are not made with excellent strategy.

We believe AI can be of assistance in this situation. However, let’s analyze the issue at hand. To begin, it’s important to recognize that tactics (and even business models) are, at their core, ideas, or consistent narratives, about how the world functions. A business strategy is basically an idea or plan for generating revenue for an enterprise. Second, the search for explanations for patterns in data is inherently inductive, therefore all theories are of this nature.

When we combine these two thoughts, we obtain a third: In order to find patterns in data and zero in on the ones that might lead to better strategic decisions, you need employ machine learning, the newest and most successful kind of artificial intelligence.

While it’s understandable that executives might think their businesses are one of a kind, it doesn’t mean they’re completely insulated from market trends and influences. A bird’s-eye view of bigger trends that individual managers may never identify if they rely just on their own experiences might be provided through machine-learning-based analysis of massive samples of aggregate industry data. Thankfully, modern strategists have access to vast troves of data on topics including as mergers and acquisitions, alliances, patents, and employee evaluations.

In a piece now out in the Strategic Management Journal, we outline the potential for this to be successful. With the use of big data and algorithms, we provide a blueprint for how strategists might find novel explanations for old problems, such as business partnerships, by employing a method called algorithm-supported induction.

Changes in private equity business models

The private equity (PE) sector relies heavily on multi-partner business alliances, sometimes known as transaction syndicates. Private equity (PE) firms, banks, and corporations (sometimes referred to as “deal syndicates”) pool their resources to acquire target companies. A private equity company’s alliance portfolio consists of the several deal syndicates that make up the business.

Around a decade ago, an innovative business strategy known as “add-on transactions” began to gain traction in this sector. In these types of transactions, a private equity firm buys a company and then combines it with another, similar company in its portfolio. Under the more common leveraged buyout approach, a private equity fund uses debt financing to purchase a standalone corporation.

We set out to answer a basic question: If private equity firms started engaging in add-on acquisitions, would their alliance portfolios need to change?

From 1990 to 2016, we analyzed around 60,000 PE agreements conducted throughout the world by more than 4,500 PE companies. We did a double-division of the sample. In the first sample, our machine learning algorithms identified two consistent trends in the uptake of complementary offers: For one, add-on deals frequently involve a previously unheard-of kind of co-investor: corporate businesses with competence in add-on agreements. For another, these co-investors are not a part of the pool of existing corporate firms that participate in leveraged buyouts. Although they are only correlations, they are rather strong and highly unlikely to be the result of chance alone.

Why would these associations exist, if at all? Creativity from the human mind is needed here. We anticipated that PE firms with a large number of existing partners, or those that already worked with companies that might compete with the desired new corporate partners, would be sluggish to adopt add-on transactions because deal syndicates need new corporate partners to pursue add-on deals.

We put these hypotheses to the test in the remaining half of our data and found solid evidence in their favor. We concluded that an organization’s preexisting portfolio of alliances has the potential to slow the implementation of novel business strategies that rely heavily on the formation of new alliances. These findings imply that there is a shadowy aspect to all partnerships.

Think like a thinker and plan ahead

Our research has important implications for strategists looking to learn more about selecting the best partners. In alliance-intensive sectors, firms need a different sort of partner for different kinds of activities, therefore business strategists need to give serious consideration to their company’s capacity restrictions and the management of rivalries among its existing partners. Another option to consider if you don’t want to share your knowledge is to hire experts.

These are hypotheses that need to be tested, maybe through smaller-scale pilot programs. Yet, they are founded on consistent trends in the collected data, making them more likely to be accurate than the opinions of managers with limited exposure and subjective experiences.

Read more: How to Use Science to Enhance the Customer Experience

Our method may find further use elsewhere, including in the PE sector. Algorithmic induction, for instance, may be used to shed light on the reasons behind the prevalence of some types of syndicated partnerships over others, given the large amount of data we have on this topic. It’s the same as wondering why certain mergers and acquisitions succeed while others fail to materialize.

Machine learning can only find correlational patterns. Yet with deliberate consideration of cause and effect, a strategist should be able to make meaningful conclusions on plausible explanations, i.e. construct a theory, which can lead to a sound strategy. Nothing is more practical than a good theory, the psychologist Kurt Lewin once stated.