Chapter 1
When do managers create value — and when are they just cost?
In January 2024, two of the world's largest employers made opposite bets on the same job.
In Leverkusen, Germany, Bayer's new CEO, Bill Anderson, faced a company in crisis: crushing debt, tens of thousands of lawsuits, expiring drug patents, and a share price in freefall. His diagnosis was unusual. The problem, he argued, was not the strategy or the products — it was the management itself. Bayer had twelve to thirteen layers of hierarchy between the CEO and the customer. Decisions crawled upward through approval chains and came back down weeks later. Anderson announced a program called Dynamic Shared Ownership: Bayer would cut its organizational layers roughly in half, eliminate about two thirds of its management positions, and hand 95 percent of day-to-day decisions to some 2,000 self-organizing teams working in 90-day cycles. Roughly 12,000 positions — most of them managers — would disappear.1 Bayer's own description of the shift is blunt: leaders would move from a "span of control" to a "span of coaching."2
That same month, in Houston, Walmart U.S. CEO John Furner stood in front of the company's annual kickoff meeting and announced the opposite. Walmart was raising the average salary of its store managers from $117,000 to $128,000, adding annual bonuses of up to 200 percent of salary, and granting Supercenter managers $20,000 a year in company stock — a package that could exceed $400,000 for a top performer.3 "A Walmart store manager is running a multimillion-dollar business and managing hundreds of people," Furner said. "We ask our managers to own their roles and act like owners. And now, they'll literally be owners."4 A year later Walmart doubled down again, lifting pay for the market managers who oversee groups of stores to as much as $620,000.16
Two sophisticated companies, looking at the same role in the same month, reached contradictory conclusions. Bayer concluded that managers are largely overhead — a costly layer that slows decisions and separates workers from customers. Walmart concluded that managers are its most leveraged asset — worth paying like executives because one good manager multiplies the output of hundreds of people. Both cannot be entirely right, at least not everywhere at once. A thoughtful observer might reasonably take either side: perhaps hierarchy really is a relic that information technology has made obsolete; or perhaps the coordination, judgment, and motivation that managers supply is invisible precisely until it is removed. This is not a rhetorical puzzle. Billions of dollars, hundreds of thousands of careers, and the daily experience of millions of employees ride on the answer.
This chapter treats that disagreement as its central question: when do managers create value, and when do they merely add cost? To answer it, you need tools, not slogans. We first define what management is (Fayol's four functions), then examine what managers actually do all day (Mintzberg's roles), how the job differs by level and skill (Katz), and the uncomfortable evidence that the activities that get managers promoted are not the activities that make their teams effective (Luthans). We then confront the empirical question directly — do managers measurably matter? — before returning to Bayer and Walmart as competing tests of the theories. By the end, you should be able to argue both sides of the January 2024 bet, and to say which conditions favor each.
Management is the process of getting work done through and with other people by planning, organizing, leading, and controlling organizational resources. The framework originates with Henri Fayol, a French mining executive whose 1916 treatise (translated into English in 1949) first argued that management is a distinct, learnable activity rather than a personality trait.5
• Planning — setting goals and deciding how to reach them: forecasting, strategy, resource commitments.
• Organizing — arranging tasks, people, and authority: who does what, who reports to whom, how units fit together.
• Leading — directing and motivating people: communicating, resolving conflict, energizing effort.
• Controlling — monitoring performance against goals and correcting course: metrics, budgets, feedback.
The four functions (often abbreviated POLC) remain the skeleton of nearly every management textbook because they describe what any organization must accomplish, whether or not a person with the title "manager" does it. That is also the framework's limit: Fayol describes management from the balcony, as an orderly cycle. It says nothing about whether the functions must be concentrated in dedicated managers or can be distributed to teams — which is precisely the question Bayer's experiment raises. Notice that Bayer did not abolish planning, organizing, leading, or controlling; it redistributed them to 90-day team cycles and coaching roles. The functions survived the managers. Whether they are performed well without them is an empirical question, not a definitional one.
A managerial role is an organized set of behaviors expected of the person holding a managerial position. Henry Mintzberg introduced the concept after directly shadowing five chief executives, minute by minute, for one week each — one of the first studies to observe what managers do rather than ask what they should do.6
Mintzberg grouped ten roles into three families:
• Interpersonal roles — figurehead, leader, liaison: the manager as the human connective tissue of the unit.
• Informational roles — monitor, disseminator, spokesperson: the manager as the unit's nerve center, collecting and routing information others do not have.
• Decisional roles — entrepreneur, disturbance handler, resource allocator, negotiator: the manager as the point where commitments are made.
Mintzberg's own data demolished the image of the reflective planner. Half of his executives' activities lasted less than nine minutes; the work was fragmented, verbal, interrupt-driven, and reactive. In a famous follow-up essay he called the orderly Fayolian picture "folklore": managers do plan, but planning happens in fragments, inside conversations, not in quiet contemplation.7 Subsequent observational studies across levels and industries have broadly replicated the fragmentation finding, though the mix of roles varies by level, function, and culture — the evidence here is descriptive and based on small observational samples, so it tells us what the job looks like, not what makes it effective.
The informational roles explain why simply deleting managers can misfire: if the manager was the unit's nerve center, removing the node scatters the information. It also explains why AI is now genuinely disruptive to the job — dashboards, copilots, and workflow tools attack the informational roles first, since monitoring and disseminating are exactly what software does cheaply. A 2025 Harvard Business School analysis of more than 50,000 software developers using generative AI found individual contributors absorbing coordination and information tasks that once defined middle management, flattening hierarchies from below.13
Mintzberg tells you what managers do, not whether doing it creates value — a busy manager and a valuable manager are not the same thing. The ten roles were derived from five American CEOs in the late 1960s; frontline supervisors, matrix managers, and remote-team leads inhabit different mixes, and the framework offers little guidance on which roles to shed when technology can perform some of them.
Organizations traditionally distinguish first-line managers (supervising non-managers), middle managers (linking strategy to operations), and top managers (setting direction for the whole enterprise). Robert Katz argued that effectiveness at each level depends on a shifting mix of three learnable skills.8
• Technical skill — proficiency in the specific work of the unit; most critical for first-line managers, who must train, evaluate, and troubleshoot.
• Human skill — the ability to work with, motivate, and understand people; roughly equally critical at every level.
• Conceptual skill — the ability to see the organization as a whole and reason about abstractions and systems; most critical at the top.
Katz explains a pattern you will see repeatedly in this course: the star engineer promoted into management and floundering. The promotion changed which skills the job rewards, and nothing in engineering built the human skills the new job runs on. It also illuminates the two cases ahead: Walmart's store manager is a first-line and middle manager fused into one — technical retail knowledge plus the human skill to run hundreds of people — which is why Walmart notes that 75 percent of its managers started as hourly associates.4 The framework's weakness is its comfortable vagueness: "human skill" spans everything from listening to political maneuvering, and Katz offers little evidence about how much each skill contributes to measurable performance.
Fred Luthans and colleagues observed 44 managers in detail (within a larger study of hundreds) and asked a subversive question: are the managers who get promoted fastest (successful managers) the same ones whose units perform best with committed, satisfied employees (effective managers)?9
Largely, no. The fastest-rising managers spent the largest share of their time on networking — socializing, politicking, interacting with outsiders — while the most effective managers spent the largest share on communication and human resource activities: exchanging information with their teams, motivating, managing conflict, developing people. Fewer than 10 percent of the observed managers ranked high on both success and effectiveness. The study is decades old, observational, and based on a modest sample in a particular era of American corporations, so treat the exact percentages as suggestive. But the core finding — that organizations often promote on visibility rather than on team results — has proven durably resonant and is consistent with later evidence that firms systematically promote their best individual performers rather than their best potential managers.
Luthans supplies the mechanism behind bad-manager horror stories: the promotion system itself selects for the wrong activities. It reframes Walmart's bet — paying store managers on their store's results, in stock, is an attempt to make success and effectiveness the same thing. The caution: the successful/effective split can be overdrawn. Networking is not waste; Mintzberg's liaison role is real work, and a manager who never manages upward cannot get their team resources.
Start with the cleanest study available. Economists Edward Lazear, Kathryn Shaw, and Christopher Stanton analyzed daily productivity data from a large technology-based services firm where workers rotate among supervisors, allowing boss effects to be separated statistically from worker effects. The findings: bosses vary enormously in quality; replacing a bottom-10-percent boss with a top-10-percent boss raises a nine-person team's output more than adding a tenth worker would; the average boss contributes about 1.75 times the productivity of the average worker; and workers assigned to better bosses are less likely to quit.10 Because it is a single firm with unusually measurable output, generalization requires care — but within its setting, the identification is strong.
Gallup's survey research points the same direction at scale: across decades of data covering millions of employees, Gallup estimates that the manager accounts for at least 70 percent of the variance in team engagement across business units.11 This figure is correlational — engaged managers may be given better teams, and both may reflect something else — so it should be read as an upper-bound headline, not a causal estimate. Still, the pattern that team-level attitudes cluster strongly by manager replicates across industries and countries, and Gallup's global data show what is at stake: worldwide engagement fell to 21 percent in 2024, at an estimated cost of $438 billion in lost productivity.12
Taken together, the evidence is strong that who manages matters. It does not follow that more management matters — and this distinction is the hinge of the whole chapter. Every layer of hierarchy adds coordination costs, approval delays, and information distortion; each additional manager is only worth it if their marginal contribution exceeds those costs. The research supports investing heavily in manager quality while remaining agnostic, even skeptical, about manager quantity. Bayer is betting against quantity; Walmart is betting on quality. The studies above are compatible with both bets.
Span of control is the number of direct reports per manager; delayering is the removal of hierarchical levels, which mechanically widens spans. Classical writers treated narrow spans (five to seven) as a principle of good administration; a century of practice has treated the "right" span as contingent on task routineness, employee skill, and information technology.
The current delayering wave is real and technology-linked. The HBS generative-AI study cited earlier finds AI absorbing precisely the reporting, coordination, and monitoring tasks that justified narrow spans, enabling flatter structures — but its authors emphasize that firms must decide deliberately which managerial tasks to automate and which human-centric ones (coaching, development, judgment under ambiguity) to reinvest in.13 Bayer illustrates the resulting arithmetic: average spans rose from single digits toward fourteen and beyond as layers were removed.15 The open question is whether the human roles that remain — the ones Gallup's and Lazear's evidence says drive engagement and productivity — can be performed well at those widths. A manager of thirty can be reached; a manager of thirty cannot coach everyone weekly.
Much of the delayering evidence is recent, consultant-generated, or self-reported by the companies doing the delayering — sources with an interest in the story ending well. Treat claims that "AI replaces middle management" and claims that "managers matter more than ever" with equal suspicion; both are currently ahead of the peer-reviewed evidence.
These two cases were chosen because they are recent, richly documented, and make opposite predictions from the same theories. If managers are primarily Mintzbergian nerve centers whose informational work software can now do, Bayer should get faster and better as layers vanish. If managers are primarily what Lazear and Gallup measure — multipliers of frontline productivity and engagement — then Walmart's premium on manager quality should outperform, and Bayer should feel the loss in coordination and morale. Each case is a test, not an illustration.
The facts: beginning in January 2024, Bayer cut organizational layers from roughly twelve to six or seven, eliminated about half its management layers and two thirds of management roles, moved some 100,000 employees toward roughly 2,000 self-directed teams on 90-day cycles, and pushed 95 percent of decisions to the teams doing the work.12 Early reported wins are concrete: a hemophilia self-infusion solution developed in one 90-day cycle instead of a projected two years; consumer products moving from insight to market in nine months; Bayer's chief transformation officer reports faster decisions, improved revenues and profitability, and more engaged teams.15 The costs are equally concrete: about 5,500 management positions eliminated, sustained internal anxiety as managers-turned-coaches adjust, and — one year in — continued financial struggles, leading skeptics to note that the reorganization has not yet cured what ails the company.14 Supporters reply that Bayer's debt and litigation predate DSO and that the model needs years, not quarters.
As a theory test: Bayer is running Fayol without dedicated Fayolians — planning via 90-day cycles, controlling via team-owned metrics, leading via coaches. Mintzberg's informational roles have been partially replaced by shared data platforms. What DSO cannot easily replace is Katz's human skill applied one-on-one: with spans stretching past fourteen, coaching capacity is the scarcest resource in the system.15
The facts: in January 2024 Walmart raised store manager average salary to $128,000, set bonuses up to 200 percent of salary, and added annual stock grants up to $20,000, taking top total compensation past $400,000.3 In January 2025 it extended the logic upward: market managers — who each oversee roughly a dozen stores — saw base pay rise to $160,000 with stock grants up to $100,000 and total packages reaching $620,000.1617 Walmart employs more than 400 market managers over about 4,600 U.S. stores; 75 percent of salaried managers began as hourly associates.4 The design is unmistakably Luthansian: compensation is tied to the manager's own store results, aligning "success" (pay, advancement) with "effectiveness" (unit performance), and the stock converts managers into the owners Furner invoked.
As a theory test: Walmart is betting that Lazear's boss effects are large in retail — that the difference between a great and mediocre manager of a store doing hundreds of millions in revenue dwarfs a few hundred thousand dollars in pay. It is also a bet on Gallup's mechanism: engagement in a high-turnover industry is built store by store, manager by manager. The risk side: Walmart is adding cost to a fixed structure. If AI-enabled tools soon perform much of the store manager's informational work, Walmart may find itself paying executive compensation for a role whose scope is shrinking.
| Bayer (DSO) | Walmart | |
| The bet | Managerial hierarchy is a cost; teams closest to customers decide best | Manager quality is the highest-leverage asset in the company |
| Core move | Cut layers ~50%, management roles ~2/3; 2,000 self-directed teams; 90-day cycles | Raise manager pay to $128K–$620K with stock and results-based bonuses |
| Theory relied on | Mintzberg (informational roles now automatable); distributed POLC | Lazear (boss effects), Gallup (engagement), Luthans (align success with effectiveness) |
| Early evidence | Faster cycle times, reported profitability gains; morale strain, unresolved financial crisis | Retention and pipeline strengthened; costly, and unproven against AI-driven role shrinkage |
| Fails if… | Coaching at wide spans cannot sustain engagement and development | The informational core of the manager's job is automated out from under the pay package |
Notice what the cases do not show. Neither is a controlled experiment; both companies changed many things at once, in different industries, from different starting points. Bayer's hierarchy was extreme (twelve-plus layers) — delayering from twelve to six may prove the evils of excess hierarchy, not of hierarchy. Walmart's stores are unusually measurable — results-based manager pay works best where results are attributable, which is precisely Lazear's setting. The honest synthesis of the evidence is conditional: managers create value where frontline work is variable, human-intensive, and measurable at the unit level (Walmart's world), and destroy value where layers exist mainly to relay information that systems can now relay (the world Bayer believed it was living in). Both companies may be right — about themselves.
Return to the central question: when do managers create value, and when do they merely add cost? The chapter's tools assemble into an answer with four parts.
First, separate the functions from the role. Planning, organizing, leading, and controlling must happen; a dedicated manager is one way to make them happen, and Bayer is a live test of the alternatives. When you evaluate any "get rid of managers" proposal, ask where each function will go — not whether it will be needed.
Second, expect the informational core of the job to keep shrinking. The roles Mintzberg found consuming managers' days — monitoring, disseminating, coordinating — are the roles software and AI perform first. What remains is what Katz called human skill and what Gallup measures as engagement-building: the parts of the job that were always hardest to do and easiest to skip.13
Third, quality beats quantity. The best-identified evidence we have says the difference between good and bad bosses is enormous, while saying nothing in favor of more bosses.10 Organizations that act on this — promoting for effectiveness rather than visibility, paying for unit results, cutting layers while investing in the managers who remain — are following the evidence. Organizations that merely cut, or merely pay, are following fashion.
Fourth, hold your conclusion loosely. Bayer's experiment and Walmart's wager are still running. By the end of this course you will have the vocabulary — motivation, teams, power, culture, structure, change — to reread both cases and notice things invisible to you now. That, and not a definition of POLC, is what this week is for.
Notes appear as superscript numbers in the text and correspond to the numbered sources above. DOIs are provided where available; classic books and pre-digital articles are cited to their original publishers.