Portolan Operations decision desk
Open Portolan
All guides
Guide · Continuous improvement and change adoption

Why process improvements do not stick, and what the research says to do about it

Improvements fade for predictable reasons, and few of them are about effort. Improvement pays off with a delay, so early poor results invite pressure on the people doing the work, which recreates the problem. Change also fails when urgency, ownership and early visible wins are missing, and when success is declared before the change is anchored. The playbook's answer: test in one area, build local ownership, expect a dip, and sustain with evidence.

Published · Updated

Results arrive late, and pressure makes it worse

Process-improvement investments pay off with a delay. Managers often read the early poor results as a sign of worker effort or a bad fit, and respond with pressure. Modelling work shows how those responses interact with how workers react and with the existing controls, so the behaviour that blocks improvement is reproduced: a self-confirming trap. The practical point is to expect the dip and to look at the system before leaning on people.

This mechanism comes from formal modelling and case-based simulation. The details vary by organisation, and the model describes how a trap can form, not whether yours has.

A checklist of how change efforts fail

Transformation efforts commonly fail through a recognisable set of errors: too little urgency, a weak guiding group, an unclear vision, poor communication, obstacles left in place, no short-term wins, victory declared too early, and changes never anchored in how the organisation works. The sequence matters because later stages depend on the credibility earned earlier. It is a diagnostic checklist drawn from cases, not a proven causal sequence.

Economic case and capability together

Change efforts tend to follow either a top-down economic approach, often fast, or a capability approach built on participation over the long term. Firms that combine both, with a clear economic case while building capability and involving people, tend to do better than those that pursue either alone. The evidence is the authors' case and consulting experience, not a controlled comparison.

Capability develops in stages, and people have to be safe to speak up

Continuous improvement can be modelled as a capability that develops in stages, from occasional projects to routine, employee-led problem solving, rather than arriving with a programme launch. In a study of 51 work teams in one manufacturing company, psychological safety, the shared belief that speaking up is safe, predicted learning behaviour, which in turn predicted team performance. That study is correlational, so it shows a relationship and not a recipe for creating safety.

Structured, participatory kaizen work in two cluster-randomised studies, in a postal service and a hospital, was associated with better employee-related outcomes. Those studies did not establish production or quality gains, and they were in two Swedish organisations.

The path, step by step

These steps come from the Continuous improvement and change adoption playbook. Each has a checkpoint you can observe.

  1. Run a structured test in one area

    Use a plan-do-study-act cycle on a small scope: predict the effect, run it for a defined period, compare with the prediction and decide to keep, adjust or stop.

    CheckpointA written prediction and a measured result from a bounded test.

    Common failureRolling out a change everywhere before testing it anywhere.

  2. Build participation and local ownership

    Give the people doing the work a routine for raising problems, proposing tests and seeing what happened. Protect the time and make the backlog visible.

    CheckpointA visible improvement backlog, with participation measured by team.

    Common failureAsking for ideas through a suggestion box and never reporting back.

  3. Manage adoption and expect the dip

    Expect performance to fall before it rises. Tell people the dip is expected, watch the leading indicators, and avoid adding pressure that drives the old workaround back.

    CheckpointThe dip is visible on a chart, with a pre-agreed response rule instead of a panic.

    Common failureAbandoning the change after the first bad week and confirming that improvement never works here.

  4. Sustain and scale with evidence

    Scale only after the test shows the mechanism working. Keep the measurement, transfer ownership to the line, and prune practices that stop earning their keep.

    CheckpointThe practice survives a change of sponsor or a busy period, with the measure still being used.

    Common failureScaling a pilot that only worked because of the person running it.

What this guide does not settle

Most of this evidence is modelling, cases or correlation. The library has no controlled trial showing that one method keeps improvements from fading in a given operation, and this guide quotes no failure-rate statistic because none of its sources supports a reliable one. Use the steps as a way to test, in one area, what holds in yours.

What the research found

Each finding below is quoted from Portolan's research library, with the limits the library records for it. Applying a result from another setting, such as a hospital or a factory, to yours is an analogy, not a proof.

Repenning & Sterman · 2002Established

Capability Traps and Self-Confirming Attribution Errors in the Dynamics of Process Improvement

Process-improvement investments pay off with a delay, but managers often attribute early poor results to worker effort or programme fit and respond with pressure. Those responses interact with workers' reactions and the existing control structure to reproduce the very behaviour that blocks improvement—a self-confirming capability trap.

Where it stops. The mechanism is developed through formal modelling and case-based simulation; specific trap dynamics vary by organisation and the model is not a predictive tool.

Kotter · 1995Applied

Leading Change: Why Transformation Efforts Fail

Transformation efforts commonly fail through a predictable set of errors: insufficient urgency, a weak guiding coalition, an unclear vision, poor communication, obstacles left in place, missing short-term wins, premature declaration of victory and failure to anchor changes in culture. The sequence matters because later stages depend on earlier credibility.

Where it stops. The model is based on case observation and has been criticised for oversimplifying messy, non-linear change; it is a diagnostic checklist, not a proven causal sequence.

Beer & Nohria · 2000Applied

Cracking the Code of Change

Change efforts tend to follow either Theory E (economic value, top-down, often rapid) or Theory O (organisational capability, participation, long-term). Firms that combine both—setting a clear economic case while building capability and involving people—tend to outperform those that pursue either pure approach.

Where it stops. The framework draws on the authors' consulting and case experience; the evidence is not a controlled comparison of change approaches.

Bessant et al. · 2001Applied

An evolutionary model of continuous improvement behaviour

Continuous improvement is modelled as an evolutionary capability: organisations move from occasional projects to routine, employee-led problem solving as behaviours and supporting routines become embedded. The case evidence shows that capability develops in stages rather than arriving with a programme launch.

Where it stops. Case-based research rather than a controlled comparison. The staged model describes a direction of travel, not a guaranteed sequence.

Edmondson · 1999Established

Psychological Safety and Learning Behavior in Work Teams

In a study of 51 work teams in a manufacturing company, psychological safety—the shared belief that it is safe to speak up and take interpersonal risks—predicted team learning behaviour, which in turn predicted team performance. Safety was not associated with lower standards or comfort; it enabled the learning that performance required.

Where it stops. The study is cross-sectional and correlational; it established the relationship between safety, learning and performance, not a simple causal recipe for creating safety.

Schwarz et al. · 2017Experimental

Using kaizen to improve employee well-being: Results from two organizational intervention studies

Two cluster-randomised organisational interventions found that structured, participatory kaizen work was associated with improved employee-related outcomes in a postal service and a hospital. The studies did not establish production or quality gains.

Where it stops. Two Swedish organisational settings with employee-focused outcomes. The mechanism may not transfer unchanged to small teams or high-turnover operations.