Integration Is a Discovery Problem

In 2023, exploration and production companies spent $234 billion on mergers and acquisitions, the most in real terms since 2012, and two deals, ExxonMobil–Pioneer and Chevron–Hess, accounted for much of it. In 2024, ten companies accounted for nearly 90% of the reserve acquisition spending of the 158 publicly traded producers worldwide that the EIA tracks. The stated rationale was operational: acreage that complements an existing position, which “can lead to lower costs and greater production efficiency”.

Executives expected more. In December 2023, 77% of 122 oil and gas executives surveyed by the Dallas Fed expected more deals of $50 billion or more within two years. In June 2024, 54.6% of 130 expected continued consolidation to lower U.S. oil production, while every executive from a producer of 100,000 barrels a day or more answered that it would have no impact.

None of the Dallas Fed’s quarterly questions since then has asked how the integration itself goes. The surveys measure deal volume and production. What happens inside the combined operation, after closing, is mostly invisible to the public record.

The most cited claim about acquisitions comes from a 2011 Harvard Business Review article: “study after study puts the failure rate of mergers and acquisitions somewhere between 70% and 90%.” The passage cites no study and does not define failure: announcement returns, profitability, synergies, or divestiture.

The meta-analytic evidence says something more modest and more useful. Across the variables most commonly studied, acquiring firms’ performance “does not positively change as a function of their acquisition activity, and is negatively affected to a modest extent”, and unidentified variables may explain significant variance in what happens after the deal. One candidate is well documented. In a case survey of 61 mergers and acquisitions that measured success as synergy realization rather than stock returns, organizational integration “was the single most important factor”.

Most acquirers do not publish synergy estimates at all. In a sample of 1,990 deals, 345 did, and the main deterrent the authors identified was the “lack of precise information on synergy values available to bidding firm management”. At announcement, management often does not know.

The recent upstream megadeals published theirs, and report beating them. ConocoPhillips announced $500 million of synergies from Marathon Oil and reported more than $1 billion on a run-rate basis in 2025. ExxonMobil’s expected Pioneer synergies went from about $2 billion a year to $4 billion, as a ten-year average, two thirds of it from improved resource recovery. Diamondback reported that it was ahead of schedule on the $550 million it promised for Endeavor.

Diamondback also said where part of the difference came from. Within the first week after closing, it had onboarded more than 1,000 employees and begun working as one organization. In the same letter to stockholders it wrote that combining two teams of basin experts was “a synergy that could not be modeled in our spreadsheet when the deal was announced”, and later described “un-modeled synergies”: a standardized facility design built from both companies’ best practices, expected to save about 10%, and better drill-out efficiency “courtesy of the legacy Endeavor team”. The value was real. It was not in the deal model, because the deal model did not know how the other team worked.

The same gap can subtract. When EQT acquired Rice Energy in 2017, the deal case included $1.9 billion in well cost synergies. A year later, EQT told investors that 2018 well costs would be over $1,000 per lateral foot rather than the $900 to $915 it had expected, raised well development capital by $300 million, and attributed most of it to the pace of activity and the learning curve on ultra-long laterals; its CFO added that the Rice synergies had not contemplated wells longer than 14,000 or 15,000 feet. Rice’s founders, in the proxy fight that followed, gave a different account: that EQT had set aside the operating blueprint the two teams spent months discussing and moved forward with its own systems, without the people responsible for Rice’s results. The two accounts come from interested parties and do not agree. They are, in the vocabulary of the previous paper in this series, a positional disagreement, and both locate the problem in how the work was done.

Even the industry’s most experienced integrator says the gap can last. Announcing the Pioneer deal, ExxonMobil’s chief executive said that XTO, acquired in 2010, “had for a long time been held somewhat separate up in Fort Worth”, and that the company had not been getting the benefits of the whole organization.

The contracts define what moves. The industry’s model operating agreement requires a former operator to deliver to its successor “all records and data relating to the operations”. Each joint operating agreement carries the COPAS accounting procedure of its vintage, and bills remain open to audit for 24 months after the end of the year in which they were rendered. In Texas, a change of operator is filed lease by lease, regulatory responsibility does not transfer until the commission approves it, and the hydrogen sulfide compliance certificate “is not transferable”. Records, data, and filings move. How the work gets done does not have a clause.

Szulanski studied 122 transfers of best practice inside eight companies and found that the major barriers were knowledge-related, not motivational: the recipient’s lack of absorptive capacity (canonical weight 0.54), causal ambiguity (0.34), and an arduous relationship between source and recipient (0.33). Lack of motivation weighed 0.05 at the source and 0.18 at the recipient. Causal ambiguity is present, he wrote, when “the precise reasons for success or failure in replicating a capability in a new setting cannot be determined even ex post”. Two caveats matter: the canonical analysis used 87 of 271 observations, and the transfers were within one firm, not between two that had just merged.

Kogut and Zander put the underlying idea in one sentence: “organizations know more than what their contracts can say.” Their empirical follow-up, on 35 Swedish innovations, found that the degree of codification and how easily capabilities are taught significantly affect the speed of transfer.

One tension should be stated rather than smoothed over. Larsson and Finkelstein found that mergers built on combining similar operations, which is most upstream consolidation, tended to elicit more employee resistance. That is a motivational account, the kind Szulanski found secondary. Both can be true. The evidence does not settle their relative weight in oil and gas.

The strongest objection is the evidence above. The largest recent deals met or doubled their targets within months, and serial acquirers buying similar targets learn: in 449 acquisitions, Haleblian and Finkelstein found a U-shaped relationship between experience and performance, and the more similar a target was to prior targets, the better the deal performed. Upstream targets are about as similar as targets get.

Two things should be conceded. Two thirds of ExxonMobil’s expected Pioneer synergies come from resource recovery, which is subsurface technology, not process. And large serial acquirers with dedicated integration teams do integrate well. The claim of this paper is narrower: it concerns the operating and administrative layer, and the mid-size operators and service companies that do not publish synergy targets at all.

The answer is in what predicts success. In 228 U.S. bank acquisitions, Zollo and Singh found that “knowledge codification strongly and positively influences acquisition performance, while experience accumulation does not”. Experience can also transfer the wrong way: in 96 organizations, second acquisitions underperformed first ones, consistent with routines from a prior deal being applied where they do not fit. Imposing the acquirer’s systems on a team whose results depended on its own is the mechanism Rice’s founders described.

If the value of an integration depends on knowledge the deal model cannot see, the first months after closing are a discovery exercise, whether or not anyone calls it that.

The literature on knowledge transfer and integration comes from manufacturing, high technology, and banking, not oil and gas. Its connection to operational discovery in upstream integration is our argument.

The synergy figures are reported by the companies themselves, as expected values or run rates, and are not audited. Firms choose when to disclose them, which biases the public record toward deals that are going well.

The EQT–Rice case rests on an earnings call transcript and on a proxy statement written by an interested party. We report both accounts and do not adjudicate between them.

We found no published measurement of how long operational integration takes in upstream oil and gas.

The public record shows that integration value depends on knowledge the deal model does not contain. It does not show how much.

Plexo builds evidence-based operational diagnostics, with a focus on oil and gas operators and service companies. AI interviewers talk with people in the field and in the office about specific, recent episodes of their work, and every claim in the resulting process map traces to a verbatim quote, with disagreements between teams reported as findings.

We intend to publish measurements from integrations only under a stated methodology, with customer consent, and on a sample large enough to support the claim. This is the third paper in a monthly series on evidence in operational diagnostics.