One statistic shows up in more M&A pitch decks than any other: somewhere between 70% and 90% of mergers and acquisitions fail. It appears in board presentations, keynote openers, and the first paragraph of one of the most widely read Harvard Business Review articles ever published on the subject. Before that number went anywhere near a page we build on M&A or financial advisory, it had to resolve to a document we had actually read, not a figure everyone repeats because everyone else already repeats it.
It resolves two different ways, and neither is the tidy story the number implies. The article most responsible for spreading it names no source at all. A real, dated, methodologically transparent study does sit behind the number people usually mean when they cite it, and reading that study directly turns up something its own citers never mention: on the study's own numbers, the failure rate is either 83% or 53%, depending on which of two definitions in the same report you use.
What we checked, and how
For each claim, the test was the same one we apply to any statistic before it reaches a page: find the actual document, not an article about the document; read it directly rather than accept a paraphrase; and quote the exact sentence. Where a publisher's page could not be reached (a repeated timeout or an HTTP response consistent with an automated-access block), that source is recorded as unread and abandoned on first refusal, with no retries and no attempt to route around the block. No paid tool, trial, or subscription was used. Total spend: zero.
The tally
| Figure as commonly stated | Verdict |
|---|---|
| "Study after study puts the failure rate of mergers and acquisitions somewhere between 70% and 90%" (Harvard Business Review, 2011) | The article names no study. Read directly, in full, with no footnote or citation attached to the sentence. |
| "83% of mergers failed to add shareholder value" | Verified. A real, named, dated survey exists, with a stated sample and a defined benchmark for success. |
| That 83% describes M&A broadly, today | Overstated. The survey is from 1999 and covers only the largest cross-border deals of 1996 to 1998, a narrow and dated slice of all M&A activity. |
| The same report supports a different failure rate | True. Elsewhere in the identical report, a different definition of success puts the failure rate at 53%, not 83%. |
The sentence everyone repeats, and the citation it doesn't have
The most-cited source for the 70% to 90% range is Harvard Business Review's 2011 article "The Big Idea: The New M&A Playbook," by Clayton M. Christensen, Richard Alton, Curtis Rising, and Andrew Waldeck. We read it directly. It opens by noting that "companies spend more than $2 trillion on acquisitions every year," then states: "Yet study after study puts the failure rate of mergers and acquisitions somewhere between 70% and 90%." No footnote follows. No study is named anywhere in the piece, then or later. The sentence functions exactly like a citation while carrying none of the information a citation is supposed to carry: which studies, what years, what definition of failure.
That is not a small gap in an otherwise minor article. It is the opening claim of one of the most influential pieces ever published on M&A strategy, and it has been repeated, unmodified, in board materials and consulting pitches for a decade and a half on the strength of "study after study" alone.
The number that actually traces to something
If a specific figure is going to be attached to that range, the one most often invoked is 83%, and it does trace to a real document: a 1999 global research report on M&A integration, produced jointly by KPMG's Transaction Services and Merger and Acquisition Integration practices. We downloaded and read the full report. Its own words: "The survey found that 82% of respondents believed the major deal they had been involved in had been a success... When we measured each one against our independent benchmark... we found that only 17% of deals had added value to the combined company, 30% produced no discernible difference, and as many as 53% actually destroyed value. In other words, 83% of mergers were unsuccessful in producing any business benefit as regards shareholder value."
That is a genuine finding, not a manufactured one. The report describes real fieldwork: interviews conducted by Taylor Nelson Sofres Harris in June 1999 with board-level respondents from 107 participating companies, drawn from a sample frame of the top 700 cross-border deals by value completed between 1996 and 1998. Success was benchmarked objectively rather than taken on the respondents' word: each deal's equity price trend, measured roughly a year after completion, was compared against the trend in its own industry, using data supplied by Bloomberg. Self-reported confidence (82% called their own deal a success) and the measured outcome (83% failed to add value) point in almost exactly opposite directions, which is the kind of gap a benchmark exists to catch.
A second number, from the same report, that most citers never mention
The report's methodology section contains a sentence that changes the arithmetic: "For standardisation purposes, we have given the benefit of the doubt to companies who had neither created nor destroyed value at the time of our research, and included those in our category of 'success'... those acquirers who had destroyed value were treated as failures." Under that definition, the one KPMG itself used for its "keys to success" cross-tabulation later in the same report, success is 17% (added value) plus 30% (no discernible difference), and failure is 53% (destroyed value only), not 83%.
Same fieldwork, same 107 companies, same report. Two internally consistent, honestly stated definitions of failure, 30 percentage points apart. A reader who quotes "83%, per KPMG" is quoting one definition; a reader who checked the whole document would also find "53%, per KPMG" sitting a few pages later. Neither is a misquote. The number was never as singular as the headline version makes it sound, even inside the one document usually credited with producing it.
What a 1999 cross-border sample cannot tell you about a 2026 deal
The KPMG report says plainly what it measured: "We focused our sample on the largest cross-border deals completed between 1996 and 1998 so the results capture the experiences of leading blue chip companies." That is one specific, dated, high-value cohort, not a cross-section of mergers and acquisitions in general. Most M&A activity, then and now, is mid-market and often domestic, a different population with different integration dynamics than the largest cross-border combinations of the late 1990s. The Harvard Business Review sentence collapses a defensible, narrow, 26-year-old finding into an undated, universal law of deal-making, and the collapse is where the number's credibility actually breaks.
What we could not verify, and left that way
A widely shared Forbes column repeating the 83% figure returned an HTTP 403 response on the first request, consistent with an automated-access block. Per the same policy applied to every source in this piece, that is recorded as unread rather than summarized secondhand; its restatement of the KPMG number is not represented here as independently checked.
McKinsey is the other name frequently attached to M&A failure-rate claims in circulation, more often in slide decks than in anything McKinsey has published. Two direct attempts to reach McKinsey's own M&A research pages returned repeated timeouts rather than loading, a pattern consistent with an automated-access block rather than a content issue. No claim attributed to McKinsey appears in this piece, because no McKinsey page was actually read.
What this means before hiring for M&A work
The number a firm leads with is a reasonable proxy for how it treats the rest of its numbers. "70% to 90% fail, study after study" is checkable in the time it takes to open the article, and the citation simply is not there. "83% of the largest cross-border deals from 1996 to 1998 failed to add shareholder value, per KPMG's own benchmark" is also checkable, and it holds up, with a named sample, a defined methodology, and a scope that has to travel with the number rather than get dropped from it.
That is the useful test for evaluating either an M&A advisory engagement or the financial advisory work behind a deal thesis: ask which document a number comes from, and whether the person citing it can also tell you what it does not cover. The integration side of the same problem, getting two organizations onto one operating model after the deal closes, is where a change management engagement earns its scope. Tell us what you're evaluating if you want the numbers behind a specific deal or proposal checked the same way.
How to check a number like this in five minutes
- Ask for the document, not the name. "Study after study" is not a citation. Neither is a firm's name with no title, year, or link attached.
- Open the document and search it for the number. The Harvard Business Review article settles this test in under a minute: the 70% to 90% range appears once, unfootnoted, and nowhere else in the piece.
- Check what "success" or "failure" was defined as. The same KPMG report produces 83% or 53% depending on which of its own two definitions is applied. That is not a rounding difference; it changes which side of "most deals" the finding lands on.
- Note the year and the sample. A 1999 survey of the largest cross-border deals from 1996 to 1998 is a specific, dated, checkable population, not a stand-in for M&A in general.
- Write down what you couldn't read. A source blocked from automated access is a different state from a source that disagrees with you, and it belongs on the record as unread rather than quietly repeated as if it had been checked.
This piece checked three claims tied to the commonly repeated M&A failure-rate statistic. It is not a survey of M&A outcomes generally, and no rate of success or failure across all mergers and acquisitions is asserted here beyond the specific, named, dated study quoted above.