The current wave of flatter-structure experiments arrives with the same confidence as the last one. Layoffs have thinned the management ranks, hybrid work has loosened the daily grip of supervision, and a new generation of founders talks about "bossless" teams as if the only thing standing between a company and greatness were a layer of managers. The honest evidence is more useful than the hype, because it shows self-management failing in patterns you can predict and, sometimes, avoid.
The case everyone cites and no one finishes
For a decade the standard exhibit for radical self-management was Zappos and its adoption of holacracy. In 2016 it was, by a wide margin, the largest company to have implemented the system in full (Bernstein, Bunch, Canner and Lee, 2016). What gets quoted less often is how the story ends. In March 2015 the chief executive gave staff an ultimatum to commit fully to the new system or take a buyout, and eighteen percent of the workforce walked (Groth, 2020). The company then quietly retreated. It cut the system's most bureaucratic meetings, brought managers back, and by 2017 had replaced holacracy with an internal market in which teams operate as small businesses with their own profit-and-loss statements and trade with one another (Groth, 2020). The poster child for going bossless did not so much prove the model as illustrate its costs.
The lesson is not that self-management is a fad. It is that a single dramatic conversion, imposed from the top under threat of exit, is close to the worst way to attempt it. Michael Lee and Amy Edmondson define self-managing organizations by a hard feature: they decentralize authority radically, formally, and systematically, eliminating the manager-to-subordinate reporting relationship across the whole organization (Lee and Edmondson, 2017). That is a structural change to how work is coordinated, not a motivational campaign, and it fails when treated as the latter.
The conditions that predict trouble
The clearest map of when self-management holds and when it buckles comes from the home-care organization Buurtzorg, which runs hundreds of self-managing nursing teams. In their study of how it scaled, the researchers name the boundary conditions plainly. The model works best where interdependence between teams is low, where frontline staff are highly motivated and competent, and where the service can be delivered by a small, relatively self-contained team (Martela and Nandram, 2025). Reverse any of those and the design starts to strain. Work that demands tight coordination across many teams still needs somewhere for that coordination to happen, and if you have abolished the managers you had better have built the mechanism that replaces them.
This fits what the broader literature already knew about hierarchy. Managerial hierarchy is genuinely good at one thing: reliable execution of known tasks in stable environments (Lee and Edmondson, 2017). Its weaknesses show up elsewhere, in complex non-routine problems that cross functional boundaries, where it tends toward rigidity and reinforces status differences that stunt people's development. So the question is never whether hierarchy is good or bad in the abstract. It is whether your particular work looks like reliable execution of known tasks or like messy cross-boundary problem-solving. Get that reading wrong and you will either flatten a structure that was doing useful work or preserve one that is suffocating it.
Why the gains are not automatic
Even when the conditions look right, self-management does not hand you a performance improvement. A systematic review of self-managing teams found that despite their wide adoption, the research is inconsistent about whether they actually raise performance, and that outcomes depend on getting the right input factors at the individual, team, and organizational levels (Magpili and Pazos, 2018). "Remove the boss and productivity rises" is not a finding the evidence supports. What the evidence supports is more demanding and more honest: performance depends on whether you have assembled the competencies, the coordination, and the emergent states of trust that a team needs to govern itself.
That reframes the Zappos retreat. The problem was not that employees could not handle freedom. It was that a large, interdependent retail operation was asked to swap its coordination system overnight, without the boundary conditions or the input factors in place, and the business metrics suffered enough to force a partial reversal (Groth, 2020). The later move to internal markets with real profit-and-loss accountability was, in effect, a search for a coordination mechanism to do the work the abolished hierarchy had been doing.
The practical read
Before flattening anything, run two diagnostics. First, measure interdependence. If the work naturally decomposes into small teams that rarely need to synchronize, self-management has a real chance. If everything touches everything, you are not removing overhead, you are relocating it onto people who now have to coordinate in their spare time. Second, measure competence and motivation on the front line, because the model assumes both and punishes their absence (Martela and Nandram, 2025). Self-management is a design choice with preconditions, and the organizations that treat it that way tend to keep it. The ones that treat it as a belief system tend, like Zappos, to end up quietly bringing the managers back.
Sources
- Bernstein, Bunch, Canner and Lee (2016). Beyond the Holacracy Hype. Establishes that Zappos was the largest full holacracy implementation and that the debate is polarized, framing the honest-assessment stance. https://hbr.org/2016/07/beyond-the-holacracy-hype
- Groth (2020). Zappos has quietly backed away from holacracy. Documents the 2015 ultimatum, the 18% buyout, and the 2017 shift to internal markets, the definitive account of the retreat. https://qz.com/work/1776841/zappos-has-quietly-backed-away-from-holacracy
- Lee and Edmondson (2017). Self-managing organizations: Exploring the limits of less-hierarchical organizing. Provides the canonical SMO definition and the account of where hierarchy helps and where it fails. https://www.sciencedirect.com/science/article/abs/pii/S0191308517300059
- Martela and Nandram (2025). Buurtzorg: scaling up an organization with hundreds of self-managing teams but no middle managers. Names the boundary conditions (low interdependence, motivated competent frontline) that predict where self-management holds. https://link.springer.com/article/10.1007/s41469-024-00184-y
- Magpili and Pazos (2018). Self-Managing Team Performance: A Systematic Review of Multilevel Input Factors. Shows the performance evidence is inconsistent and depends on input factors, undercutting the automatic-gain claim. https://doi.org/10.1177/1046496417710500