The "flat versus hierarchical" framing of organizational structure has become a marketing category. Tech firms in particular have spent fifteen years selling the idea that flat structures produce more innovation, more employee satisfaction, and faster decision-making than traditional hierarchies. The empirical record is messier. Lindred Greer's organizational-behavior research at Michigan and Stanford has demonstrated that strict status-hierarchy compression can produce more conflict and slower decisions in certain settings. Bob Sutton and Huggy Rao's Scaling Up Excellence (2014) documented multiple cases where "flat" organizations re-developed shadow hierarchies that were less legible and accountable than the explicit ones they replaced. Frederic Laloux's Reinventing Organizations (2014) inspired a wave of self-management experiments — Zappos's holacracy adoption being the most famous — that largely produced mixed-to-negative outcomes and quiet retreats. Valve's famous "no managers" structure has been reported by former employees as having produced an opaque seniority system rather than a flat one.
The argument here is that "hierarchical versus flat" is the wrong question. The right question is which decision-rights structures, accountability mechanisms, and management norms produce the outcomes a given organization needs. The flat-organization rhetoric got popular because it sounded like worker empowerment. The empirical record suggests that what actually drives the outcomes the rhetoric promised — innovation, autonomy, fast decisions, retention — is psychological safety, clear decision rights, and accountable managers, regardless of how flat the org chart is.
What the empirical literature says about flat structures
Three findings worth holding onto.
First, the relationship between organizational structure and innovation is not what the flat-org rhetoric assumes. Research by Daniel Levinthal, Daniel Albert, and others on organizational ambidexterity suggests that decentralized exploration combined with centralized exploitation outperforms either pure structure in most settings. The "ambidextrous organization" framing developed by Charles O'Reilly and Michael Tushman implies that the structure question depends on the specific innovation challenge, not on a universal preference for flatness.
Second, the holacracy and self-management experiments have produced limited evidence of sustained success at scale. Zappos's adoption, the most studied case, produced documented attrition spikes, manager exodus, and eventual quiet de-emphasis. Tony Hsieh's death in 2020 and Zappos's subsequent management evolution have moved the firm away from strict holacracy. Other self-management firms (Buurtzorg in Dutch home healthcare, Morning Star in tomato processing, Patagonia in retail) have made it work in specific industry contexts but have not produced replicable templates for general adoption.
Third, the strongest predictor of team performance in the empirical literature is psychological safety — Amy Edmondson's framework, validated in Google's Project Aristotle and dozens of subsequent replications. Psychological safety can be designed into hierarchical organizations and can be absent in flat ones. The structure question is downstream of the management-norms question.
What "flat" usually means in practice
The "flat organization" framing covers at least four distinct structural designs, and the empirical record varies sharply across them.
Reduced layers
Cutting middle-management layers in firms that have accumulated too many. The Drucker Institute and Charles Handy's organizational research support this in specific cases — firms that have grown layered through acquisition or risk-aversion can produce meaningful productivity gains by removing layers. The decision is structural and finite, not philosophical.
Wider spans of control
Increasing the number of direct reports per manager. The empirical record on span of control, including HBR research and the Conference Board's surveys, suggests that effective span varies sharply by role complexity — wider for routine work, narrower for complex knowledge work. The "flat = wide spans" version of flat organizations tends to overshoot, producing managers with too many reports to coach effectively.
Self-managed teams or holacracy
Removing the manager role entirely for some or all units. The empirical record is mixed-to-negative outside specific industry contexts. The Buurtzorg case in Dutch home healthcare is the strongest success, but the conditions that made it work (small autonomous teams, mature professional workforce, defined service offerings) do not generalize.
Dual-track ladders
Senior individual contributor tracks parallel to management tracks, with equivalent compensation. This is increasingly standard in tech and has the strongest empirical support of any "flat-adjacent" intervention. Google, Microsoft, and most large tech firms have implemented versions; the implementation quality varies but the model is well-validated. Will Larson's writing on engineering ladders is the most accessible practitioner treatment.
What actually drives the outcomes the flat-org rhetoric promised
The outcomes that flat-organization advocates pitch — fast decisions, innovation, autonomy, retention — track to specific management norms more than to structural design.
Psychological safety
The Edmondson framework, validated in Google's Project Aristotle research on 180+ teams, identifies psychological safety as the strongest predictor of team performance. Building psychological safety is a manager-behavior challenge, not a structural one.
Clear decision rights
The RAPID framework (Recommend, Agree, Perform, Input, Decide), developed by Bain & Company and popularized in the management literature, structures who decides what without requiring structural flatness. Firms with clear decision rights operate faster than firms with ambiguous ones, regardless of how many layers separate top from bottom.
Manager quality
The Gallup Q12 surveys and the Google "Project Oxygen" research have consistently identified manager quality as the single strongest predictor of team engagement and retention. The implications are that investing in manager training, removing low-performing managers, and structuring promotion criteria around demonstrated management capability produce the outcomes that "flat" structures promise.
Information transparency
Firms with transparent information flow — public dashboards, shared metrics, accessible strategy documents — operate faster than firms with information hoarding, regardless of organizational structure. GitLab's all-public handbook, Stripe's internal openness norms, and Bridgewater's principles-based culture each implement this through different mechanisms.
The hierarchical version, taken seriously
Hierarchical organizations are not inherently bureaucratic, slow, or stifling. The U.S. military, large industrial firms, and many of the world's most operationally excellent companies are deeply hierarchical and produce outcomes that flat-organization advocates would credit to flatness. The Hierarchical-versus-flat framing obscures the actual operational variables.
The honest version is that hierarchy is a tool with specific uses: it enables clear accountability, structured coordination at scale, and reliable execution of routine work. The Aspen Institute Job Quality Center's work, John Kotter's organizational-change research, and the broader management literature all support the view that well-designed hierarchies in stable-process settings outperform flat alternatives.
The flat-organization rhetoric usefully challenges complacent hierarchies. It does not provide a universal replacement.
The leadership implication
The leadership-development question that follows from this is more interesting than the structural-design question. What does management look like in organizations that are trying to combine hierarchical accountability with the autonomy, psychological safety, and decision-rights clarity that flat-organization rhetoric promised?
For the full treatment, see our flagship Leadership in the Augmented Workplace →, which examines how leadership roles are evolving as AI tools change what management work involves.
"Flat versus hierarchical" is the wrong question. The right question is which decision-rights, accountability mechanisms, and management norms produce the outcomes you need. Most of the gains the flat-organization rhetoric promised come from psychological safety and clear decision rights, which can exist at any altitude on the org chart.
The organizational-structure conversation in management literature has cycled through "lean," "agile," "matrix," "holacracy," "self-management," "team of teams," and now "AI-native organization" in roughly fifteen-year waves. Each cycle produces useful insights and overpromises. The honest accumulated wisdom from a half-century of organizational research is that structure is downstream of management norms, that the outcomes leaders want are mostly driven by psychological safety, decision rights, and manager quality, and that the universal templates are universal mostly in failing to generalize. The next wave of organizational redesign — driven by AI-augmented productivity, distributed work, and the changing nature of management work itself — will produce its own templates and its own overpromises. The empirical center will continue to be the same things it has been for a long time.
Updated May 21, 2026. This piece was substantively rewritten as part of NWLB's 2026 editorial refresh.
