Organizational change doesn’t stick when the new way of working depends on temporary attention while everyday incentives, authority, resources, and routines continue supporting the old one. Under pressure, people may return to familiar practices because those practices remain easier, safer, or more rewarded. Sometimes, however, reversing a change is a sensible response to evidence that it is not working.
Your transformation moved backward. Before restarting it, establish what that backward movement means. A learning dip, a useful correction, repeated oscillation, and genuine regression require different responses. Treating all four as resistance can make the problem worse.
This guide explains why improvements unravel, how to distinguish setbacks from deterioration, and how to sustain useful change through System Shaping. The aim is to improve outcomes and preserve the ability to adapt—not to defend every part of the original plan.
Why organizational change doesn’t stick
A rollout changes what is formally available. Sustained change requires people to use it successfully in ordinary work. Training attendance, a new organization chart, and a completed implementation checklist do not establish that the new approach is workable or producing benefits.
Buchanan and colleagues’ review of sustaining organizational change identifies interacting influences across multiple levels, including leadership, resources, organizational arrangements, politics, and context. That supports investigating several conditions together rather than assuming one universal cause of backsliding.
The new behavior conflicts with real rewards
A company asks teams to share expertise but rewards only individual output. Sharing becomes optional work that competes with what affects recognition and progression. Repeating the collaboration message will not resolve that contradiction.
Responsibility changes without usable authority
Teams become accountable for faster decisions but still need several approvals to act. Under deadline pressure, informal escalation returns. The formal operating model and the actual decision process remain different.
Adoption consumes capacity that nobody protects
Learning, migration, troubleshooting, and parallel workflows take time. If existing commitments remain untouched, employees may restore old shortcuts to keep services running. Examine change saturation before interpreting workarounds as unwillingness.
Temporary support conceals an unfinished transition
A pilot succeeds with dedicated experts, exceptions, and daily leadership attention. Wider adoption removes that support. The problem may be a missing operating capability rather than a decline in commitment. Plan who owns maintenance, onboarding, coaching, and exception handling after the program ends.
These mechanisms can interact. Limited capacity can increase errors; errors can trigger extra approvals; approval delays can encourage workarounds. For a broader assessment of failed initiatives, see why organizational change fails. Here, the focus is what happens after apparent progress.
Why system evolution is not a straight line
Nonlinear organizational change means that effort, timing, and results do not necessarily move in a steady proportion or sequence. Several small adjustments may produce little visible effect until a constraint is removed. One team may improve while another struggles with the dependencies created by that improvement.
A 12-year study by Amis, Slack, and Hinings examined the pace, sequence, and linearity of radical change. Its findings describe nonlinear change within organizations and emphasize the importance of sequencing particular changes. It does not establish a universal cycle that every organization must follow.
In this article, system evolution means changes in organizational patterns over time. It does not imply inevitable advancement, biological evolution, or a ladder of human worth. “Forward” must refer to something observable: safer operations, more reliable delivery, better service, or an improved ability to respond to changing conditions.
Back-and-forth movement can also leave something behind. A team may temporarily restore an approval step while retaining better information, stronger relationships, and clearer limits. Its process resembles an earlier arrangement, but its capabilities are not necessarily identical. Conversely, new terminology can conceal the return of old constraints.
A plateau may provide time to consolidate useful practices. It may also indicate stagnation. The distinction depends on evidence, not the shape of a diagram. Stability and transformation can support one another when stable commitments make learning possible.
Is this a setback, a correction, or genuine regression?
Start by describing the observed change: which behavior or outcome moved, for whom, compared with what baseline, and over what period? A decline in one dashboard measure does not explain its cause. Check whether demand, reporting, staffing, or the definition of the measure changed.
| Pattern | Evidence to examine | Possible response |
|---|---|---|
| Learning dip | Performance falls during adoption while proficiency develops. | Protect learning time and set a review point. |
| Useful reversal | The new approach produces unacceptable problems or no longer fits the context. | Modify or reverse the affected element. |
| Repeated oscillation | Similar triggers repeatedly produce the same gains and reversals. | Investigate feedback, incentives, and unresolved trade-offs. |
| Genuine regression | Useful practices disappear and intended outcomes deteriorate. | Revisit authority, support, capacity, and rewards. |
| Uneven progress | Local gains create difficulties for other teams or dependencies. | Coordinate changes and assess overall outcomes. |
More than one pattern can be present. A team can experience a genuine learning dip while a neighboring function restores an unnecessary gate. Avoid assigning a reassuring label too early. If proficiency is not improving, support is insufficient, or harm exceeds the agreed limit, “learning takes time” is not an adequate explanation.
Likewise, low adoption does not prove regression if the new process is unsuitable. Ask whether people are avoiding an effective practice or correcting an ineffective one. Investigating their reasons is more useful than assuming resistance to change explains everything.
Why old patterns return under pressure
Pressure changes the consequences of decisions. A leader may encourage experimentation when delivery is comfortable, then punish the first visible error when a deadline approaches. Employees receive a clear signal about which commitment carries more weight.
Consider a possible reinforcing loop: an incident prompts additional approvals; approvals increase waiting; waiting compresses execution time; compressed work encourages shortcuts; further incidents appear to justify more approvals. Each response can seem reasonable locally while contributing to the wider problem.
This is a hypothesis to test, not a diagnosis to impose. Inspect decision waiting times, incident causes, workload, and exceptions. Approval requirements may be necessary and effective; the issue is whether their scope and consequences match the risk.
Feedback delays complicate judgment. A practice may take time to influence outcomes, while its implementation costs appear immediately. But a delayed benefit is not a license to extend an unsuccessful initiative indefinitely. Define what intermediate evidence should appear and when the decision will be revisited.
When the same response keeps recreating the same difficulty, recurring system conditions deserve attention. “The system returns to its old pattern” is shorthand for interacting decisions and constraints, not a claim that an organization has a mind or intention of its own.
Worked example: autonomy followed by a return to approvals
The following scenario is illustrative, not a reported company case or evidence of typical results. A delivery organization delegates routine release decisions to teams. The goal is to reduce waiting without weakening reliability. Teams receive decision rights, but risk thresholds and cross-team dependencies remain unclear.
After a release disrupts a shared service, executives restore central approval for every release. A review labels the change a failure. Yet the incident alone cannot establish whether delegation was the cause: the failure could involve missing dependency information, insufficient testing, unclear boundaries, or another mechanism.
Response A: indefinite centralization
Every release requires approval, with no review date. This may provide oversight, but it can also create waiting and transfer ownership away from teams. Examine whether the gate actually prevents the relevant failures and whether its costs are justified. Centralization is not automatically regression; unexplained loss of capability and worsening outcomes would strengthen that diagnosis.
Response B: bounded containment
Approval applies temporarily to releases affecting the shared service. An accountable owner reviews the incident, maintains a dependency register, and sets a date to reassess the restriction. Exit conditions specify the evidence needed to remove or narrow the gate. This makes the temporary control a testable response rather than a permanent default.
Response C: revised delegation
Teams retain routine release authority within explicit risk limits. Decisions affecting shared services require consultation or escalation. The organization improves dependency information and rehearses exception handling. This preserves autonomy where it is useful while addressing a specific weakness.
Compare the responses against the same questions: Did relevant incidents decrease? Did decision delays increase? Are people better able to recognize and escalate risk? What happens when the usual expert is absent? Choose based on the evidence and operating context, not a general preference for centralization or decentralization.
How to respond when organizational change moves backward
1. Establish what actually reversed
Write a specific observation: “Routine decisions now wait for senior approval” is more useful than “The culture is regressing.” Separate changes in behavior, capability, outcomes, and reporting. Name the affected teams and the period being examined.
2. Revisit the intended benefit
Confirm that the original outcome still matters and that the intervention remains plausible. A change designed for a different workload or risk environment may need revision. Define the boundaries that must be protected while you investigate, including service continuity and acceptable risk.
3. Test competing explanations
Ask people doing the work what became harder, riskier, or less supported. Compare their accounts with observed workflows and operational evidence. Look for counterexamples: if one team sustains the practice, what differs in its authority, workload, expertise, or dependencies? Do not assume a single successful team proves the approach will scale.
4. Contain harm without freezing the whole organization
When immediate containment is necessary, define its scope, owner, review date, and exit conditions. Apply controls to the relevant risk rather than extending them automatically to unrelated work. Preserve records and capabilities that will help the organization learn from the incident.
5. Change the conditions sustaining the problem
Select a bounded intervention connected to the diagnosis. This might mean removing conflicting targets, funding ongoing support, clarifying a decision boundary, or reducing simultaneous initiatives. Communication should explain the changed conditions and expected behavior; it cannot substitute for making the behavior feasible.
6. Review and choose: continue, adapt, reverse, or stop
Agree in advance what evidence would support each decision. Review often enough to detect harm, but allow an appropriate period for the intended mechanism to work. Record what was learned even when the intervention stops. Continuing unchanged should require a reason just as reversal does.
How to measure whether change is becoming durable
Durability means useful practices and benefits persist for an appropriate period and can be maintained with realistic operating support. It does not mean removing every resource, preventing further adaptation, or waiting for a crisis to test people.
| Dimension | Useful evidence | What it does not prove |
|---|---|---|
| Adoption | Observed use in real workflows, across affected groups. | Logins or attendance alone do not establish effective use. |
| Outcomes | Relevant benefits compared with a baseline and changing demand. | A short improvement does not establish lasting impact or causation. |
| Support | Named ownership, onboarding, maintenance, and adequate capacity. | A successful pilot does not prove routine support is sufficient. |
| Pressure response | Behavior during ordinary peaks, staff absence, and exceptions. | A calm period does not show what happens under different conditions. |
| Adaptability | Evidence prompts timely, reasoned adjustments. | Frequent changes do not automatically demonstrate learning. |
Use a small set of measures that match the intended benefit. For the release example, combine decision waiting time, relevant incident severity, and observed use of risk boundaries. Review distributions and affected teams rather than relying only on an organization-wide average. Record changes in demand and reporting that could explain apparent movement.
For a broader measurement approach, see organizational transformation metrics. Do not turn “zero reversals” into a success target: it could reward hiding mistakes or persisting with a harmful design.
From sustaining change to System Shaping
System Shaping directs attention to the conditions that make behavior possible, rewarded, and repeatable. In this article, it is the interpretive framework connecting the diagnosis to intervention—not a claim that the five diagnostic categories have been scientifically validated.
Instead of asking only how to restore adoption, ask what the return of an old practice reveals. Is authority unclear? Does a target reward the workaround? Has capacity fallen? Does important information arrive too late? The answer helps identify which condition to change and which useful stability to preserve.
Recovery can restore essential functioning without producing a higher level of capability. The Resilience Loop explores the possibility of rebuilding beyond the previous pattern; that is a possible outcome, not an obligation or guarantee. Evolutionary intelligence adds the question of whether lessons actually change future decisions.
The goal is not uninterrupted forward motion. It is the ability to recognize what is happening, preserve what remains useful, and adjust the conditions that determine the next outcome. Explore the System Shaping library for further reading and practical resources.
Frequently asked questions
Why does organizational change fail to stick?
New behavior may depend on temporary support while incentives, decision rights, capacity, and routines still favor the old approach. It may also be unsuitable for real working conditions. Investigate both the feasibility of adoption and the value of the change itself.
Is regression a normal part of organizational change?
Setbacks can occur, but they are neither inevitable nor automatically useful. A temporary learning dip differs from sustained loss of capability. Examine the trajectory, causes, and consequences before deciding whether to support, modify, or stop the approach.
How do you distinguish a learning dip from genuine regression?
Look for evidence that proficiency is increasing, problems are being resolved, and outcomes are recovering within an appropriate review period. Persistent deterioration without learning or adequate support weakens the learning-dip explanation. Several causes may operate simultaneously.
Can going back to an old process be the right decision?
Yes. Reversal can protect service, reduce harm, or correct a flawed assumption. Clarify why it is necessary, what learning will be retained, and whether the restored process is temporary or intended to remain. An older practice is not automatically inferior.
What does nonlinear organizational change mean?
It means changes in effort, timing, and outcomes need not follow a steady, proportional sequence. Delays, interactions, and different rates of adoption can produce uneven trajectories. Nonlinearity does not mean that planning is useless or that progress is guaranteed.
How can leaders sustain organizational change?
Make the intended behavior workable through clear authority, compatible incentives, sufficient capacity, ongoing support, and timely feedback. Monitor outcomes as well as adoption, and retain the ability to adapt or abandon changes that no longer serve their purpose.
Research and references
- Buchanan, D., et al. (2005). No going back: A review of the literature on sustaining organizational change. International Journal of Management Reviews, 7(3), 189–205.
- Amis, J., Slack, T., & Hinings, C. R. (2004). The pace, sequence, and linearity of radical change. Academy of Management Journal, 47(1), 15–39.
The research above informs the discussion of sustainability and change trajectories. The diagnostic comparison, review prompts, and worked example are practical synthesis developed for this article; they are not assessment instruments validated by those studies.