IBM Is Betting on Algorithms To Improve Its Acquisition Success - and Its Growth

Well, if you have to buy rather than make to grow, at least you can use something you developed to improve your chances of success.

Assuming, of course, that the algorithms you developed, are actually correct about the financial analysis, to say nothing of all that messy interpersonal stuff that happens when two organizations attempt to combine. JL

Sujeet Indap reports in the Financial Times:

After 14 consecutive quarters of declining sales IBM is turning increasingly to acquisitions in the hope of reviving growth.Using its experience of scores of deals in recent years, its software, and its hundreds of research scientists, IBM began experimenting with a computer algorithm that could spot the big risks in target companies during the M&A due diligence process.
After 14 consecutive quarters of declining sales IBM is turning increasingly to mergers and acquisitions in the hope of reviving growth. But one of its chief dealmakers is upfront about the risks acquisitions carry.

“If 70 per cent of M&A fails, would you propose spending $20bn doing deals?”, asks Paul Price, IBM’s director of M&A integration, alluding to the widely held belief about the effectiveness of dealmaking — and the sum that the company allocated to deals between 2010 and 2015.However, the technology group hopes to achieve far more than a 30 per cent success rate by stripping out what Mr Price argues is the chief cause of deals going bad: human error.

Using its experience of doing scores of deals in recent years, its vault of software, and its hundreds of research scientists, five years ago IBM began experimenting with a computer algorithm that could spot the big risks in target companies during the M&A due diligence process.

Like most large companies, IBM has a “corporate development” team that evaluates and executes transactions in conjunction with the company’s operating units. Its aim in creating a due diligence algorithm was to better align the acquisition process with post-deal integration plans. “The focus used to be on the downstream part of the process: target identification,” Mr Price says. “Integration concerns . . . couldn’t stop a deal.”
Now, according to Martin Schroeter, IBM’s chief financial officer, the M&A Pro tool has given the company a sharper insight into deal risk and the ability to complete deals faster, before competitors can pounce or a target can get cold feet.
“We have been able to dramatically speed up and improve the process of acquisitions from identification and due diligence through to integration and execution”, Mr Schroeter says.
IBM says it can now begin and finish a deal in as little as three weeks. A 2014 study from McKinsey, the consultancy, concluded that the companies that were most effective at M&A typically finished deals more quickly.

M&A Pro is constructed as a “machine learning” system: an algorithm that learns to make judgments based on historical data. Its core technology comes from SPSS, a computational and statistical software company IBM bought in 2009 for $1.2bn. M&A Pro also draws on a handful of other software companies that IBM has bought including Cognos, a financial reporting software package it acquired for $4.9bn in 2009.

More than 200 potential factors from more than 100 acquisitions have been whittled down to 28 variables. Among those that IBM was willing to disclose were the target company’s geographic scope, the region where it has most employees, and its intention to expand its software-as-a-service business model.
IBM’s scientists then refine their calculations each quarter as more data are added.

Not everyone is an M&A process expert. But what we have done is create a common language, an Esperanto, for deal execution across the organisation- Paul Price, IBM’s director of M&A integration

What M&A Pro produces from its regression analyses are visualisations of integration risks, qualitative advice and scorecards assessing the financial performance of previous deals.
Its assessments allow executives to dig into each core risk it identifies. For example, if M&A Pro reveals concerns about the sales function at a target company, it can go deeper into specifics such as sales force capacity, then deeper still into how the company hires sales people.
At the same time, M&A Pro highlights the areas where there is a chance for revenue to grow faster than the business case.
A summary forecast calculates the likelihood of IBM hitting its performance target for the deal, spitting out a conclusion such as: “Deals like Project X tend to miss the business case, averaging 81 per cent revenue attainment.”

Another set of data on post-merger integration risks can offer qualitative assessments of how a possible acquisition target differs from similar companies IBM has bought in the past — such as its development team being based offshore.
Finally, a financial dashboard shows the performance of past deals against the business plan at the time of acquisition.
Mr Price estimates that IBM ran M&A Pro 30 or 40 times in 2015, adding that the purpose was not to consolidate deal decision-making among a handful of people, but rather the opposite.
“Not everyone is an M&A process expert,” he says. “But what we have done is create a common language, an Esperanto, for deal execution across the organisation . . . Our business now is much more grounded in economic and operational reality.”
But data-driven dealmaking has not turned around IBM’s poor performance. Its total shareholder return for the past two years is minus 25 per cent. Even $20bn in M&A spent on emerging software companies will not immediately boost sales for a group with $80bn in annual revenue.

$20bnSum IBM allocated to deals between 2010 and 2015

Nevertheless, IBM appears pleased to be able to move quickly and chase multiple M&A targets simultaneously. Clients have already asked about how they could use M&A Pro. Mr Price is not aware of any similar algorithms in use but believes pharmaceuticals R&D is a natural extension for deal risk analytics.
Mr Price says IBM has “the best deal risk management in the industry”. However, he acknowledges that — even with its algorithmic approach — its dealmaking record will not be perfect. “You are still bringing together two living, breathing organisms,” he says.