A reorganization begins with a chart that looks cleaner than the company it describes.
The layers are fewer. The spans are tidier. Each function sits in a box that appears to know what it is for. After months of murky decisions and slow work, the new structure offers something almost medicinal: visible order.
Then Monday arrives.
The same decisions still circle through 4 meetings. The same 2 leaders disagree about priorities. The same team waits for work that everybody supports and nobody owns. Only now, people also have new managers, interrupted relationships, and a vocabulary full of phrases such as "operating model transformation."
An organizational reorganization can solve a structural problem. It can put related work together, separate incompatible mandates, reduce an overloaded layer, or give a neglected priority real authority. What it cannot do is convert a behavioral or operating problem into a structural one merely because structure is easier to draw.
That distinction explains a good deal of reorg disappointment. The chart changes. The system that made the old chart fail does not.
Start with the failing work, not the new chart
Most bad reorgs are not irrational. They are overconfident answers to poorly specified questions.
"We need to move faster" is not a diagnosis. Neither is "we need more accountability," "the teams are too siloed," or "we should be closer to the customer." These are descriptions of dissatisfaction. Each could point to a structural problem, but each could also point somewhere else.
Slow decisions may come from too many approval levels. They may also come from leaders who will not state a priority, a culture that punishes a wrong call more than a late one, or a funding process that requires every choice to become a small constitutional convention.
Weak accountability may come from overlapping roles. It may also come from work that has never been assigned to a named owner, success measures that conflict, or leaders who reverse decisions without acknowledging that they did so.
Silos may reflect formal boundaries that genuinely obstruct work. They may also reflect incentives that reward local results, incompatible systems, missing relationships, or a history that taught 2 groups not to trust each other.
All of those problems are real. Only some require a new org chart.
The first discipline of a reorg is therefore rather unglamorous: describe the failing work in observable terms. Which decision takes too long? Where does it stall? Which handoff loses information? Whose goals conflict? What work has no owner? Which manager has no time to manage?
Until those questions have answers, the structure is being asked to carry a theory nobody has tested.
The famous failure rate needs an honest label
The most repeated reorganization statistics tend to arrive with suspiciously precise confidence. A large percentage of reorgs fail. Productivity falls by an exact amount. Performance returns after a standard number of months. The numbers vary, but the sales pitch underneath them is remarkably stable: change is hard, so please buy a change program.
There is better evidence, but it still needs restraint.
In 2010, McKinsey reported a global survey of 2,525 executives. Of the respondents, 1,890 said their companies had undergone an organizational redesign. Among the relevant responses, 16% said the redesign had fully met its objectives in the planned time. Only 37% said the organization met its implementation timeline.
Those are bleak results. They do not prove that 84% of all reorgs fail. The study relied on executives' retrospective reports, "success" included both objectives and schedule, and different companies were attempting different changes under different conditions. It tells us that complete, on-time success was uncommon in this large survey. It does not provide a law of organizational physics.
That caveat actually makes the useful point sharper. Leaders should distrust both the fantasy that a reorg will cleanly solve a complex problem and the cottage industry of universal failure statistics built around that fantasy. The sensible position is less theatrical: reorganization is consequential, difficult to evaluate, and easy to mistake for progress before results arrive.
The survey also found that broad goals were common. Companies wanted lower costs, faster decisions, greater accountability, stronger customer focus, or higher growth. These are legitimate aims. They are also broad enough to conceal several different operating problems.
A reorg built around a slogan can look coherent on a slide while leaving the mechanism completely unspecified.
People do not simply "resist change"
When a reorg struggles, employees are often accused of resisting change. This explanation is wonderfully convenient because it locates the problem inside the people with the least control over the design.
The research is less tidy.
Shaul Oreg, Maria Vakola, and Achilles Armenakis reviewed 79 quantitative studies of how change recipients react. The studies covered mergers, downsizing, technology changes, work redesign, and other interventions across many kinds of organizations. Reactions were related to characteristics of the individual, the internal context, the change process, what the change actually altered, and whether people expected it to help or harm them.
That is not a finding that people love change if leaders communicate more. It is a finding that reactions have reasons.
Someone whose role becomes less coherent may be responding to ambiguity, not nostalgia. A manager who loses the relationships that made their work possible may be protecting a functioning network, not a title. A team asked to support a new strategy while measured against the old one has noticed an implementation defect.
Some resistance is self-interested. Some is based on incomplete information. Some is an early warning from people close enough to the work to see what the design missed. Treating all 3 as the same emotional obstacle is not change management. It is a way to avoid learning.
The useful question is not "How do we get people on board?" It is "What does this reaction tell us about the design, the process, the trade-offs, or the interests affected?"
Leaders do not have to accept every objection. They do have to understand it before deciding what it means.
A new chart cannot repair an old operating system
Structure assigns formal authority and groups work. An operating system determines how people use that structure each day.
It includes decision rights, goals, incentives, planning rhythms, information flows, staffing, meeting habits, escalation paths, and the small social permissions that tell people whether they can raise a problem without being punished for the inconvenience.
If the problem lives in that system, moving reporting lines may only move the problem.
Consider a product and sales organization that repeatedly overpromises delivery dates. The visible conflict sits between 2 functions, so a leader combines them under 1 executive. That might help if split authority caused the promises. It will do little if revenue targets still reward commitments that delivery capacity cannot support, forecast information remains poor, and nobody has the authority to reject a bad deal.
Or consider a leadership team that wants fewer escalations. It removes a management layer and widens spans of control. Decisions do move closer to the work on paper. In practice, the new managers inherit more reports, keep most of their individual work, and have less time to coach people through judgment calls. Escalations either continue or go underground.
The chart was not irrelevant in either case. It was incomplete.
Jeroen Stouten, Denise Rousseau, and David De Cremer reviewed scholarly evidence alongside popular change models. They found that widely used practitioner models often leaned heavily on expert opinion and that the research did not support a single neat sequence for every change. Their synthesis identified 10 evidence-based steps, including diagnosing the problem, developing a compelling direction, setting goals, creating supportive structures, building participation, and evaluating progress.
The important word is not "10." It is "diagnosing." A structure should follow the diagnosis and fit with the other changes required. It should not be the diagnosis wearing rectangles.
Reorgs also disturb the company nobody drew
A reporting line can be changed in a system before lunch. A trusted working relationship cannot.
Every organization depends on informal networks of advice, trust, and communication. People learn who can solve a particular problem, who remembers why a decision was made, and who can translate between groups that use the same words differently. That informal organization is a second chart, even when nobody has mapped it.
A reorg changes the conditions around those relationships. Some connections survive and help the new design function. Others fade when people no longer share meetings, work, systems, or context. New teams may inherit formal responsibilities without inheriting the knowledge network that supported them.
This is one reason a clean break can be expensive. The organization may spend months rediscovering which person knows the exception, which team owns the dependency, and which relationship allowed 2 groups to solve a problem without escalation.
The answer is not to preserve every old relationship. Some informal networks compensate for bad design, exclude newcomers, or concentrate too much dependency in one person. A reorg is an opportunity to examine them, not to romanticize them.
Before moving work, ask:
- Which advice relationships are essential to this work?
- Where is trust doing the job of a missing process?
- Which people bridge groups that will now move farther apart?
- What knowledge has no durable home?
- Which new relationships must exist for the proposed structure to work?
Then give those relationships time and a reason to form. A kickoff meeting is not a network.
When a reorg is the right tool
None of this is an argument for leaving a bad structure alone. Sometimes the boxes really are the problem.
A reorganization is more likely to be appropriate when:
- authority sits somewhere other than the consequences of the decision;
- work that must coordinate closely is separated by avoidable boundaries;
- 2 roles have genuinely incompatible mandates;
- a strategic priority has no resources, leader, or organizational home;
- a management layer adds approval without adding judgment or support;
- spans of control make the actual management work impossible;
- the company has changed enough that the old grouping no longer matches its work.
Even then, "change the structure" is not a complete intervention. The new design needs explicit decision rights, role boundaries, measures, resources, and transition work. The people affected need enough information to understand what changed, what did not, why the trade-off is worth making, and where unresolved questions will be decided.
The design also needs a falsifiable promise. If the reorg is meant to speed a particular decision, measure its time and quality before and after. If it is meant to reduce duplicate work, identify the duplication now. If it is meant to improve customer ownership, name the customer outcomes and the person who will own them.
"Better collaboration" is not a measure. It is what organizations write when they would prefer not to discover whether the reorg worked.
Make the smallest structural change that fits the diagnosis
Large reorgs feel decisive. They also change several variables at once, which makes the results harder to interpret and the transition harder to manage.
If the evidence points to 1 broken decision path, fix that path. If 2 teams need a shared owner, change that relationship. If a priority lacks capacity, move the capacity and authority it needs. Do not redraw the rest of the company to make the slide symmetrical.
This principle sounds cautious because it is. Caution is useful when the intervention changes managers, careers, relationships, identity, and the routes through which work gets done.
It also preserves the ability to learn. A focused change can be evaluated. A company-wide redesign followed by new goals, new systems, new leaders, and a new planning process produces a result, but not a clean explanation.
The org chart should make the intended structure visible. It should also leave no doubt that reporting lines are not the same as ownership. Assign the work between the boxes. State who decides when priorities conflict. Protect the management time the design assumes. Rebuild the critical connections it disrupts.
Then watch what happens in the work, not merely whether the new chart has been uploaded everywhere.
The chart is a hypothesis
A reorg is a theory about how work will improve if authority, people, and resources are arranged differently. Treat it with the seriousness a theory deserves.
State the problem before proposing the structure. Gather evidence from the work. Test other explanations. Name the mechanism by which the new design should help. Preserve the relationships worth preserving. Measure the outcome. Be willing to admit when the problem survived its new box.
Sometimes a reorg is exactly right. Sometimes the company needs clearer ownership, a repaired decision process, better goals, or managers with enough time to manage. Often it needs several of those things together.
Moving the boxes is the visible part. Making the new arrangement work is the job.
And if the proposed solution depends on each manager absorbing several more people, the next question is not whether the chart looks flatter. It is whether those managers can still do the work of managing.