Chapter 6
Leadership performed vs. leadership practiced
In October 2014, two women were running technology companies in California, and only one of them looked like a leader.
Elizabeth Holmes was everywhere: the black turtleneck, the unblinking stare, the improbably deep voice, the story polished to a gleam — a Stanford dropout at nineteen, now building a machine that would diagnose disease from a single drop of blood.3 Theranos was valued at $9 billion, making her, on paper, the youngest self-made female billionaire in history; her investors included Rupert Murdoch and the Walton family, and her board glittered with statesmen — George Shultz, Henry Kissinger, former senators and generals.23 She gave the speeches, sat for the magazine covers, and performed vision the way the leadership industry teaches it.
That same month, Lisa Su became CEO of a company almost nobody wanted. AMD's stock was trading around $1.61; the company carried roughly $2.5 billion in debt, had lost the processor war to Intel and the graphics war to NVIDIA, and was the subject of open bankruptcy speculation. Su's own mentors advised her not to take the job.56 There were no magazine covers. Her announced plan had three unglamorous parts: build great products, deepen customer trust, simplify the company.7 Then she spent years doing exactly that, betting the firm on a new chip architecture that would take until 2017 to ship.
A decade later the ledger read as follows. Holmes was in federal prison in Texas, serving 11 years and 3 months after a jury convicted her of defrauding investors of more than $140 million; the technology had never worked.123 Su had led what a leading semiconductor historian called one of the great turnarounds in modern American business: AMD passed Intel in both market value and annual revenue in 2022, grew from a $2–3 billion company to one worth hundreds of billions, and Su was named Time's CEO of the Year in 2024.5678
The easy moral — charisma bad, substance good — is too easy, and you should distrust it. The research you are about to read shows that visionary, inspirational leadership has genuine, well-replicated positive effects; charisma is not fraud, and most charismatic leaders are not Elizabeth Holmes. Meanwhile, quiet technical excellence fails all the time: plenty of brilliant engineers have run companies into the ground for lack of exactly the storytelling and coalition-building Holmes was so good at. The genuine puzzle is that our perceptual machinery for detecting leadership — presence, confidence, narrative — is only loosely connected to the machinery that produces results. In October 2014, virtually every observer using the standard cues ranked Holmes above Su. What were they measuring?
The central question: what do effective leaders actually do — and how much of "leadership" is a story we tell about the people in charge? The tools: the trait evidence and the crucial gap between leader emergence and leader effectiveness; the two master behaviors every leader mixes; contingency thinking — matching leadership to situations; transformational and transactional leadership, the field's dominant modern framework; charisma and its documented dark side; and the relationship view — LMX and the evidence on humility. Then Holmes and Su return as the decade's cleanest natural experiment in performed versus practiced leadership.
Leadership is the process of influencing others toward shared goals — with or without formal authority (which is why Week 4's manager-versus-leader distinction matters: managers hold positions; leaders hold followers). The oldest research program asked whether leaders share identifiable traits. Ralph Stogdill's famous 1948 review found no trait that guaranteed leadership across situations, temporarily burying the idea;9 the Big Five revived it, and the modern meta-analytic verdict (from Week 2's Judge and colleagues) is that extraversion, conscientiousness, and openness relate meaningfully to leadership, with a multiple correlation near .48.10
The indispensable refinement hides inside that meta-analysis: traits predict leader emergence (who gets seen as a leader and chosen) more strongly than leader effectiveness (whose groups actually perform). Extraversion, height, confidence, and fluent talk help people acquire leadership roles; they are far weaker guarantees of what happens next. This single distinction is the skeleton key to the hook — and to every hiring committee you will ever sit on: the cues that make someone look like a leader and the behaviors that make their unit succeed are correlated, but loosely. Selection systems that cannot tell the two apart systematically staff themselves with impressive emergers. The trait approach's other limits are familiar from Week 2: modest correlations, situation dependence, and nothing about what leaders should actually do — which is where research went next.
Mid-century programs at Ohio State and Michigan, watching what leaders did rather than who they were, kept finding the same two dimensions. Consideration (people-orientation): showing concern, building trust, supporting followers. Initiating structure (task-orientation): defining roles, setting standards, organizing the work. The dimensions are independent — a leader can be high on both, either, or neither — and a half-century later, Judge, Piccolo, and Ilies's meta-analysis confirmed that both matter: consideration relates more strongly to follower satisfaction and motivation, initiating structure more strongly to group performance, and the best outcomes come from leaders high on both.11 The framework's virtues are its teachability — behaviors, unlike traits, can be learned — and its durability: almost every later model, transformational leadership included, is recognizably these two dimensions wearing new vocabulary. Its limit is the one that launched the next era: the behavioral studies assumed one best style everywhere, and the data kept showing that what works depends on the situation.
Contingency theories hold that leadership effectiveness is a match between style and situation. Fred Fiedler's pioneering version argued the match runs through situational favorability: task-oriented leaders excel when situations are very favorable or very unfavorable, relationship-oriented leaders in the middle — and, provocatively, that style is fixed, so organizations should re-engineer situations rather than retrain leaders.12 Robert House's path–goal theory made the logic practical: the leader's job is to clear the follower's path to the goal, supplying whatever the situation lacks — direction for ambiguous tasks, support for stressful ones, participation for capable and unconsulted teams, challenge for the under-stretched — and leader behavior that duplicates what the situation already provides is redundant or insulting.13 The contingency era's specific models have had mixed empirical careers (Fiedler's measure in particular remains contested), but its core insight is now simply true across the field: there is no context-free best style, and the diagnostic question — what does this team, on this task, currently lack? — is the most useful sentence in this chapter for a new manager. Notice its kinship with Week 2's situation strength and Week 3's expectancy theory: leadership here is expectancy repair.
The modern era's dominant framework began with James MacGregor Burns's distinction between transactional leadership — an exchange of rewards for performance — and transforming leadership, which elevates followers' goals and identities; Bernard Bass operationalized it for organizations.1415
• Transformational — the "four I's": idealized influence (being a model worth identifying with), inspirational motivation (meaningful vision), intellectual stimulation (challenging assumptions), individualized consideration (developing each follower).
• Transactional — contingent reward (clear goals, honest exchange) and management-by-exception (intervening on deviations).
Judge and Piccolo's meta-analysis across 87 studies found transformational leadership a valid predictor of follower satisfaction, motivation, and performance (overall validity about .44) — but delivered two corrections to the hype. First, contingent reward — plain, transactional clarity about goals and rewards — performed nearly as well (about .39), and the two are strongly correlated: the best leaders do both, and transformational leadership works best on top of reliable transactional foundations, not instead of them. Second, much of the evidence is same-source and correlational: inspired followers rate their leaders as inspiring, so the causal arrow is murkier than the effect sizes look.16 The honest summary: vision and meaning add real value — after the promises are kept.
Transformational leadership's failure mode is its first component detached from the rest: vision without truthfulness. The framework itself cannot distinguish a transformational leader from a skilled performer of transformational leadership — the four I's describe behaviors an audience observes, and Holmes exhibited every one of them to investors, employees, and journalists for a decade.23 That is not a quirk; it is the framework's central vulnerability, and it is why the next section exists.
Charisma is best understood not as a substance leaders possess but as an attribution followers make — a judgment triggered by confident vision, unconventionality, personal risk-taking, and crisis. That framing explains both its power and its danger: because charisma lives in the audience's perception, it can be manufactured, and it detaches from competence. Rakesh Khurana's study of CEO labor markets documented the institutional version: boards searching for "corporate saviors" systematically overpaid for charismatic outsiders, prioritizing presence and story over operating knowledge, with no reliable performance payoff.19 The individual version is the well-documented association between narcissism and leader emergence — the confidence displays that win selection — paired with null-to-negative links to actual effectiveness, plus elevated risk-taking and resistance to bad news. None of this makes charisma toxic per se: attributed charisma amplifies whatever it is attached to, mobilizing extraordinary effort behind sound strategies and unsound ones alike. The managerial defense is procedural, not cynical: separate the evaluation of the claim from the evaluation of the claimant — exactly the separation Theranos's celebrity board, long on statesmen and short on laboratory scientists, failed to perform.3
Leader–member exchange (LMX) theory relocates leadership from the leader's traits to the dyadic relationship: leaders develop differentiated relationships with each follower, ranging from high-quality exchanges (trust, latitude, mutual obligation) to low-quality, strictly contractual ones. Gerstner and Day's meta-analysis found LMX quality among the strongest and most consistent predictors in the leadership literature — relating substantially to follower performance, satisfaction, commitment, and (negatively) turnover intentions.17 Its uncomfortable twin finding: differentiation is real, and visibly unequal in-groups corrode the justice perceptions Week 3 established. The relationship view pairs naturally with the evidence on humility. Jim Collins's study of companies that sustained transformations found them led not by celebrities but by "Level 5" leaders — a paradoxical blend of personal humility and fierce professional will, who "look out the window" to credit others for success and "in the mirror" to assign blame.18 Collins's method (selecting on outcomes, then reading leadership backward) cannot prove causation, and humility can shade into indecision; but the pattern converges with LMX and with the emergence/effectiveness gap: the behaviors that build high-quality relationships and durable results are quiet ones, systematically underweighted by every selection process tuned to detect presence.
October 2014 supplies an unusually controlled comparison: two CEOs, same state, same broad industry, same decade — one maximizing every emergence cue the leadership literature catalogs, the other exhibiting almost none of them. If leadership is what it looks like, Holmes should have built the more valuable company. If leadership is a set of practices — structure, credibility, relationships, kept promises — the comparison should invert. It inverted completely, and the theories predict exactly how.
The facts: founded in 2003 by a 19-year-old Stanford dropout, Theranos raised $945 million and reached a $9 billion valuation on the claim that its device could run comprehensive tests on a finger-prick of blood; the board featured Shultz, Kissinger, former senators, and retired military leaders.23 The technology did not work, and the company concealed it — running tests on modified commercial analyzers while telling investors otherwise. The Wall Street Journal's 2015 reporting began the unraveling; the company dissolved in 2018; and on January 3, 2022, a federal jury convicted Holmes of conspiracy and three counts of investor fraud involving more than $140 million in wire transfers (acquitting on patient-related counts). She was sentenced to 135 months, ordered with her co-defendant to pay restitution equal to investors' total investments, and reported to prison in 2023; her appeal was rejected in 2025.124
As a theory test: Holmes is the emergence/effectiveness gap run to its logical extreme. Every mechanism in this chapter's first half worked flawlessly as an emergence engine: trait cues (confidence, vision, unconventionality — the turtleneck was a deliberate citation of Jobs), all four I's performed convincingly, charisma attributions cascading from one prestigious validator to the next.319 What was missing was everything in the chapter's second half: no functioning transactional foundation (the promises exchanged for capital were false); a board constructed for halo rather than oversight — statesmen with no laboratory-science expertise to evaluate the claims;3 and an internal regime of secrecy and silenced dissent that inverted every LMX and psychological-safety prescription (whistleblowers, not lieutenants, brought the truth out). The sentencing judge's phrase — "failure by fraud" is not normal startup failure — marks the analytical line: the case does not indict vision; it indicts vision unsecured by truth, operating in a validation environment engineered never to check.1
The facts: Su took over in October 2014 with AMD near $2–3 billion in market value, roughly $2.5 billion in debt, a sub-$2 stock, and open bankruptcy speculation; mentors warned her off the job.568 Her three-part plan — great products, customer trust, simpler operations — translated into a bet-the-company investment in the Zen architecture (shipped 2017), exit from money-losing activities, and patient cultivation of exactly the customers AMD had burned.7 The compounding was slow, then sudden: by 2022 AMD had passed Intel in both market capitalization and annual revenue and completed the $49 billion Xilinx acquisition; by mid-decade its market value had multiplied by roughly a hundred, and Su — an MIT-trained engineer who frames strategy in five-to-ten-year arcs ("everything takes time") — was named Time's CEO of the Year for 2024.5678
As a theory test: Su's decade is the chapter's second half executed in order. Transactional bedrock first: the earliest wins were credibility wins — commitments to customers and engineers kept on schedule, restoring the contingent-reward machinery a decade of broken roadmaps had destroyed.716 Structure and consideration together: ruthless simplification of the portfolio alongside visible respect for engineering talent — the Ohio State pair, both high.11 Contingency logic: a demoralized technical organization lacked believable direction, so she supplied precisely that — an expert CEO whose technical fluency made the vision credible rather than merely inspiring; expert power (Week 4) doing the work charisma performs elsewhere. Transformational elements arriving last and cheapest: by the time AMD's story became inspirational, it was simply a description of results. The honest caveats: Su rode industry waves (BC cycles, then AI) that lifted her bets; Intel's own stumbles were a gift; survivorship and halo effects now gild every decision in retrospect — the same attributional machinery that built Holmes up operates, more deservedly but no less automatically, on Su.19
| Holmes / Theranos | Su / AMD | |
| Emergence cues (2014) | Maximal: story, presence, celebrity board, $9B valuation | Minimal: distressed firm, $2 stock, mentors advising against |
| Transactional foundation | Broken: promises exchanged for capital were false | Rebuilt first: roadmaps kept, customer trust restored |
| Source of influence | Attributed charisma, validated by prestige cascade | Expert power + kept commitments; vision credible because technical |
| Oversight environment | Halo board without domain expertise; internal secrecy | Engineering culture; results checkable every product cycle |
| Decade outcome | Fraud conviction; 135-month sentence; company dissolved | Passed Intel in value and revenue (2022); Time CEO of the Year 2024 |
Held carefully, the comparison teaches three things — and refuses a fourth. It confirms the emergence/effectiveness gap as the central hazard of leadership selection: every evaluator in 2014 using presence-based cues rank-ordered these two leaders exactly backward.1019 It illustrates the sequencing finding hidden in the transformational meta-analysis: inspiration compounds only on top of kept promises, and the fastest way to distinguish vision from performance of vision is to check the transactional layer underneath.16 And it shows that verification environments, not follower gullibility, are the difference that matters: Theranos engineered away every checkpoint (secrecy, a non-expert board, punished dissent), while semiconductors impose checkpoints every benchmark cycle — Su could not have faked Zen. What the comparison does not license: the conclusion that charisma predicts fraud or that quiet leaders are reliably effective. Most visionary founders are honest; plenty of humble operators fail. The lesson is procedural — evaluate claims independently of claimants — not physiognomic.
Return to the central question: what do effective leaders actually do, and how much of "leadership" is a story we tell about the people in charge? Four conclusions hold.
Separate the audition from the job. Traits and presence predict who gets chosen far better than what happens next.910 When selecting leaders — or becoming one — weight kept commitments, unit results, and follower testimony over the confidence display, because the display is exactly what the market for corporate saviors overpays for.19
Do both master behaviors, in the right order. Structure and consideration are the irreducible pair;11 transactional clarity is the foundation and transformational meaning the multiplier — a .39 floor before a .44 ceiling, never the reverse.16
Lead by subtraction: supply what the situation lacks. Direction for ambiguity, support for stress, autonomy for competence.13 The question "what does this team need that it doesn't have?" will serve you longer than any style you could adopt.
Build relationships and verification at the same time. High-quality exchanges are the strongest relational predictor we have,17 and humility — crediting outward, blaming inward — is what keeps a leader auditable.18 The leaders worth following construct environments where their own claims can be checked; the ones to escape construct environments where they can't. That single test would have sorted October 2014 correctly — and it requires no gift for reading souls, only the discipline to look under the story.
Notes appear as superscript numbers in the text and correspond to the numbered sources above. DOIs are provided where available; classic books are cited to their original publishers.