Post-merger integration: testing the sequence before rollout

A global consulting firm was advising on the integration of two New York professional-services firms: a 5,000-person acquirer and a 200-person firm it had acquired in a transaction worth hundreds of millions of dollars. The integration had to retain talent while spreading the acquired firm’s technology and working practices across the larger organisation, and the consulting team had one attempt at it. Besides the recommended sequence, our case study reports three findings from Artificial Societies’ simulation:

1. 1.

   Sequencing mattered more than messaging. Strategies that began with technology rollout before partner alignment produced the largest retention declines among senior people in the acquired firm, even when the message itself tested well.

2. 2.

   Roughly 40 mid-level employees whose employment histories connected the two firms held no formal authority, yet adoption moved fastest in the simulations where they were engaged first.

3. 3.

   Firm-wide announcements met more resistance than the same message from a partner the recipient already knew.

Acquired startup employees leave faster than ordinary hires. In Daniel Kim’s study of about 4,000 US high-tech startup acquisitions between 1990 and 2011, 33% of acquired workers left within the first year, against 12% of regular hires with similar skills and experience, MIT Sloan reported (opens in a new tab) in January 2019. Those were startup employees rather than professional-services partners, so read the rates as a reference point, not a forecast for your deal.

Bain’s practitioner research puts the people side at the centre of failure. Nearly half of respondents to its M&A Practitioners’ 2023 Outlook Survey named cultural fit, or difficulty integrating management teams, as a primary reason their past deals failed (Bain, January 2023 (opens in a new tab)). Differences between the firms matter more to the dominated side. In two surveys of merged organisations, perceived differences between the partners were more negatively related to identification with the merged organisation among members of the dominated firm than among members of the dominant one (van Knippenberg and colleagues, 2002 (opens in a new tab)).

The consulting firm in our post-merger case study faced three constraints: testing a rollout sequence on real teams would change the integration before the comparison finished, informal influence sat outside the organisation chart, and internal communications data was difficult to obtain.

Rather than simulate all 5,000 people at the acquirer, Artificial Societies modelled the 1,500 across the combined organisation with direct or second-order working relationships with the acquired team. We built their personas from public social-interaction and employment-history data, using real, anonymised professional profiles.

Artificial Societies then created independent copies of that network and tested 12 strategies in parallel, each across four intervention phases, tracking how each sequence affected talent retention and technology adoption. The consulting team could follow every strategy through the integration, locate the point at which it began to fail and refine the sequence before recommending it to the client.

Retention lists drawn up by senior leaders can overlook key people, because complex hierarchies and unconscious bias hide who really matters, McKinsey’s 2020 guidance on M&A talent retention warns. Its bottom-up alternative of interviews, surveys and social network analysis is not always feasible before close. David Krackhardt and Jeffrey Hanson argued in Harvard Business Review in 1993 that informal networks can restart stalled initiatives, or block communication and organise opposition to change.

Artificial Societies’ post-merger simulation found roughly 40 mid-level employees whose employment histories connected the two firms. None held formal authority, and adoption moved fastest in the simulations that engaged them first. Employment histories show where people have worked, not whom they trust today, so check a list like this with people inside both firms before acting on it.

The messenger changed the reaction in Artificial Societies’ post-merger simulation. Firm-wide announcements produced persona responses that were variations on “this is being done to us, not with us”, while the same message met less resistance when it came from a partner the recipient already knew. That line is simulated persona speech, not an employee’s words, and in our case study responses like it explained the findings. On retention more broadly, a McKinsey survey of more than 1,400 integration executives found praise and commendation from an immediate manager to be the most effective lever (October 2020). Our article on testing corporate strategy announcements covers the wording.

An integration simulation will not show you every relationship that matters. Artificial Societies built the post-merger network from public professional data, so it maps informal influence imperfectly: a falling-out between two teams may leave no public trace. A retention decline among personas is a warning about a sequence, not a count of resignations, and the roughly 40 connectors the simulation found matter only if someone then talks to them.

The accuracy figure we publish comes from a survey benchmark: Artificial Societies’ distribution accuracy, the overlap between simulated and human opinion distributions, is 86% across 1,000 surveys, against a 91% ceiling set by humans answering the same question twice (Survey Evaluation Report, January 2026, on our method and evaluation page). It measures opinion, and it is not a measure of how well a simulation forecasts retention or adoption.

Our post-merger case study shows what an integration engagement involved. It began with the population that mattered: the 1,500 people with direct or second-order working relationships with the acquired team, not the acquirer’s full 5,000. We built their personas from public professional data. The 12 strategies, each split into four phases, then ran in parallel on independent copies of the network against talent-retention and technology-adoption targets, and the consulting team could see where each began to fail. The engagement took six weeks. Results come back as a report, with full access to the network in Radiant, our platform, to re-run the analyses. To scope a simulation for your deal, book a call.