What the research says

Why do changes stall after launch?

The short answer

In the cases we mapped, changes faded or failed once under way when old hierarchies came back, middle managers were left powerless, or bad news was filtered on its way up, though no study here measures stalling directly. In one of two hospitals, where nurse middle managers coped less well with pressure to meet regulations and save money, a change that worked at first faded as old hierarchies reasserted themselves (Spyridonidis & Currie, 2016). At a financial services firm, a failed change was traced to closed-off executive and rank-and-file systems that left middle managers powerless. At Nokia, a smartphone effort suffered as fearful middle managers shared less bad news upward and leaders grew overly optimistic. And few studies touch on reviewing a change once it is running: 6 of 173 as of 30 October 2026, none rated high strength.

Based on 12 studies from the Atlas of the Middle: 0 rated high strength, 9 medium, 3 low. Designs: qualitative case (9), qualitative interviews (2), survey (1).

The old way comes back

Two studies show earlier ways of working reasserting themselves, or persisting in memory in ways that both helped and hindered, after a change began. In a low-rated bank case, managers' choices about keeping hybrid work or returning to old practices followed what local pressures drew their limited attention to.

The middle can't act or can't speak

Cases where middle managers were left powerless, filtered bad news, or turned against their leaders.

The middle is overloaded

In five UK organizations going through large-scale restructuring, including delayering and downsizing, middle managers reported more autonomy but longer, more intense and more closely monitored work.

What happens day to day

One study of hospital units reports that day-to-day interactions between unit managers and nursing staff set off virtuous or vicious change cycles. In a low-rated case at a highly formalized bank, contradictions across levels held change back.

What keeps a change moving

Studies on what sustains or redirects a change: routines where top and middle managers review results together, tied to stronger performance in one survey; early results that build or drain commitment; and a change that continued after its champion left.

Broader evidence from the Evidence Library

These come from syntheses and trials on organizational change generally, not only middle managers, graded with the same rubric as the Atlas studies.

  • Williams et al. 2025 · cluster randomized trial (the WISDOM trial), 21 youth mental health clinics (452 youths); this paper is in Psychiatric Services, read from its abstract · rated High strength

    The implementation-leadership program (LOCI) did not sustain higher measurement-based care completion or better symptom improvement, a clear null result. Among the 81 youths who received the care, fidelity was higher at LOCI sites (24% against 1%), though low in both arms (rated Medium).

    The Atlas cases show changes fading once under way; this trial found that a leadership program's advantage did not carry into the sustainment period, in a controlled setting. It is the same trial as Williams et al. 2024 (Implementation Science) and 2026 (Translational Behavioral Medicine), so it counts once; the 2026 paper, counting clinic-level monthly uptake rather than completion and symptoms, found a smaller advantage still there at 35 months, so the papers agree that the gain shrank rather than held.

  • Becker et al. 2025 · cluster randomized trial, 28 opioid treatment programs (186 providers) · rated High strength

    Providers at programs given an enhanced strategy (incentives plus team facilitation) were more likely to reach an excellent competence benchmark, significant in two of three models, but whether programs sustained the practice did not differ (4 vs 6 programs); with 28 programs, power to detect a sustainment difference was limited, and that comparison is rated Medium.

    It adds that extra launch support may have lifted provider skill, though not in every model, while whether programs kept the practice going did not differ; the Atlas cases cannot test this, and the trial had little power to detect a difference in sustainment.

  • Hughes et al. 2020 · meta-analysis of pooled correlations on training sustainment · rated Medium strength

    Peer, supervisor and organizational support were all positively related to long-term sustainment of learned skills, with peer and supervisor support showing the strongest relationships.

    It is about keeping trained skills in use, not about a whole change, and it is correlational; it fits the Atlas group on what keeps a change moving.

Still untested in the Evidence Library (Loop Part 4): no trial of a pause or initiative freeze, and no test of review cadence. Both trials are in health care.

What I take from it

My read is that most changes don't fail at the announcement. They fade in the months after, when no one is reviewing them and the managers carrying them are stretched. That is a hypothesis the research hasn't tested.

This is my reading, Gabriel Lopez's, not a finding of the studies above.

Common questions

Why do organizational changes fail after implementation?

The studies we mapped point to several patterns rather than one cause: old hierarchies reasserting themselves, middle managers left powerless or holding back bad news, and middle managers facing longer, more intense and more closely watched work after restructuring.

How long does it take for a change to stick?

The research we mapped does not answer that. Only 6 of 173 studies (as of 30 October 2026) touch on reviewing a change or holding it steady, none is rated high strength, and none tests how long a change takes to settle or a specific timeline.

Every summary on this page was written by Unfrozen from the study's abstract or full text and checked twice against the source. Loop coding is exploratory and provisional until an independent human coder checks it. Spot an error? Email info@unfreeze.us and it gets fixed.

More questions