{"id":4389,"date":"2026-05-21T12:54:06","date_gmt":"2026-05-21T16:54:06","guid":{"rendered":"https:\/\/paulsyng.com\/blog\/?p=4389"},"modified":"2026-05-21T12:54:08","modified_gmt":"2026-05-21T16:54:08","slug":"the-costume-industry-part-iii-musk-did-not-kill-tesla-the-original-position-did-its-job","status":"publish","type":"post","link":"https:\/\/paulsyng.com\/blog\/the-costume-industry-part-iii-musk-did-not-kill-tesla-the-original-position-did-its-job\/","title":{"rendered":"The Costume Industry, Part III: Musk Did Not Kill Tesla. The Original Position Did Its Job."},"content":{"rendered":"\n

On the difference between the fame that follows proof and the fame that replaces it.<\/em><\/p>\n\n\n\n

The Harvard Business Review just published a piece<\/a> arguing that Tesla’s brand is now hostage to its CEO, and that the fix is for the company to manage Musk’s perception more carefully. The piece is built on a 2,200-respondent conjoint study run by Bryan Orme at Sawtooth Software in November 2025, and authored by Elie Ofek at HBS. The data point doing the work in the article is this: when respondents were told Musk owned Tesla, modelled demand reshuffled toward competing EV brands at a rate the authors call structurally significant.<\/p>\n\n\n\n

I want to take it seriously because the underlying observation that Tesla’s market position is weakening is correct. The diagnosis of why it’s weakening is what does not survive scrutiny.<\/p>\n\n\n\n

I have been studying Tesla’s positioning from outside the company since 2021. Customer language across Reddit, Trustpilot, YouTube long-form reviews, the resale market, owner forums, and the public filings. No inside access. No NDA. No conjoint apparatus of my own. The framework I use is the 4-Level Positioning Canvas (Frame, Execute, Live, Position) and the distinction at the center of this series: vocabulary debt versus decision debt.<\/p>\n\n\n\n

1\u2014 What the conjoint actually measured<\/h2>\n\n\n\n

A conjoint study measures stated preference under hypothetical tradeoffs. Respondents are shown bundles of attributes, told to choose between them, and the algorithm reverse-engineers an importance weight for each attribute. When you add “Elon Musk is the CEO” as an attribute and find that respondents now choose other brands, you have learned something specific: under the conditions of being asked, in 2025, with the current political weather, people say they would prefer a Tesla less.<\/p>\n\n\n\n

That is real. It is also stated preference, not revealed behaviour. It does not measure what those respondents would actually do<\/em> at a dealership with a finance offer on the table and a Model Y in front of them. It measures what they say when asked.<\/p>\n\n\n\n

The HBR piece moves from “respondents say their preference reshuffles” to “Tesla’s brand is now defined by its CEO” in two paragraphs. That is the leap that does not hold.<\/p>\n\n\n\n

The charitable read: conjoint is a useful instrument for measuring directional shifts in stated preference, and a 2,200-person sample, weighted properly, will surface a signal. Ofek and Orme are not making it up. The signal exists.<\/p>\n\n\n\n

The push: signal that someone’s stated preference has shifted is not evidence that the company’s positioning was authored by the CEO. It’s evidence that the CEO’s current visibility is being weighed against the company’s accumulated position. Those are different objects.<\/p>\n\n\n\n

2 \u2014 The causal arrow that the article reverses<\/h2>\n\n\n\n

Tesla owned the future of transportation in the customer’s mind before Musk had 200 million followers on X. That position was paid for in specific capital decisions made between roughly 2008 and 2018: the Roadster as proof-of-concept, the Model S as a luxury sedan that genuinely beat its competitors on performance and software, the Supercharger network built ahead of demand, the Gigafactory committing to vertical integration when every other automaker was outsourcing batteries, the Model 3 hitting volume in 2018 against analyst consensus that it would not.<\/p>\n\n\n\n

Each of those decisions was costly. Each one refused something. The Supercharger network refused the convention that charging was someone else’s problem. The Gigafactory refused the convention that scale came from supplier networks. The Model 3 refused the convention that Tesla was a niche luxury play.<\/p>\n\n\n\n

The noun in the customer’s mind was built decision by decision, year by year, before the persona scaled past the company.<\/p>\n\n\n\n

Musk’s fame is downstream of that position, not upstream of it. The fame compounded because the position was real. If the position had been hollow, the fame would have made the hollowness more visible, faster, and Tesla would have ended where Quibi ended.<\/p>\n\n\n\n

This is the line Part I of this series was put in writing one week ago.\u00a0Fame is the residue, not the cause.<\/strong>\u00a0I am restating it here because the HBR piece is the most recent and most credentialed version of the inversion, and the inversion is doing damage to how operators think about their own companies.<\/p>\n\n\n\n

3 \u2014 The seven layers underneath the conjoint<\/h2>\n\n\n\n

I wrote a piece in March called\u00a0The Seven Layers of Research Bias<\/a><\/em>\u00a0that laid out why traditional market research methodology produces data biased in a consistent and predictable direction \u2014 toward confirming the narrative the research was designed inside. I want to apply that frame directly to the conjoint, because it shows why a well-run study by serious researchers can still produce a wrong diagnosis.<\/p>\n\n\n\n

Conjoint analysis sits inside at least four of the seven layers.<\/p>\n\n\n\n

Layer 1, framing.<\/strong> The study asked respondents to evaluate vehicle bundles with attributes including “CEO is Elon Musk.” That framing presupposes that the CEO is a brand attribute on the same level as price, range, and warranty. The frame is the conclusion. Respondents who had not previously thought of the CEO as a purchase attribute were prompted to treat him as one, and the data the frame produced was then read as evidence that the CEO is a brand attribute. The loop closed before the first respondent answered.<\/p>\n\n\n\n

Layer 3, demand characteristics.<\/strong>\u00a0Orne’s 1962 finding holds: every research interaction signals to the respondent what the right answer is. In late 2025, with the political weather around Musk at maximum saturation, respondents shown his name as a discrete attribute will infer that the study is investigating the Musk effect. They will produce data that confirms the inference. Corneille and Lush (2023) showed the deeper version: respondents don’t just give modified answers, they form modified subjective experience. After the survey, they remembered being more bothered by Musk’s politics than they actually were before the survey.<\/p>\n\n\n\n

Layer 4, social desirability and acquiescence.<\/strong>\u00a0Saying “I would buy a Tesla less if Musk owned it” costs the respondent nothing and signals the correct social affiliation in the current cultural frame. Saying “I would still buy one” carries social cost in some networks. Stated-preference data, asked about a politically loaded CEO in 2025, will overstate the reshuffling effect against revealed behaviour by a measurable margin. Hill and Roberts (2023) found that acquiescence bias alone can inflate prevalence estimates by up to 50%.<\/p>\n\n\n\n

Layer 7, the espoused-theory gap.<\/strong> Argyris and Schon’s distinction lands hardest here. Respondents’ espoused theory (“I would not buy a Tesla because Musk”) differs systematically from their theory-in-use (what they actually do at the dealership when the lease offer is competitive and the alternative is a Mach-E). The conjoint measures espoused theory at industrial scale. It is structurally incapable of measuring theory-in-use, because asking the question contaminates the answer.<\/p>\n\n\n\n

This is not a critique of Sawtooth’s instrument or Ofek’s rigour. Conjoint is the correct tool for what conjoint is designed to do \u2014 measure stated preference under hypothetical tradeoffs. The instrument did its job. The interpretive leap, from “stated preference reshuffles when Musk is named” to “the CEO has become the brand,” is the layer the seven-layer stack makes invisible to everyone in the room.<\/p>\n\n\n\n

The data was clean. The architecture around the data was contaminated. The headline confirmed the narrative that the architecture was built inside.<\/p>\n\n\n\n

The charitable read:<\/strong> the authors are operating within a methodological tradition that is universally accepted in business research, and have produced findings consistent with that tradition’s standards. The push: the tradition itself starts inside the company’s narrative and asks people what they think, and the conclusions it generates reliably confirm whatever narrative the researcher started with. In this case, the narrative the article was built inside is that the CEO is the brand. The study found data consistent with that narrative. The data does not establish the narrative. It was generated inside it.<\/p>\n\n\n\n

If you want to know what is actually true about Tesla’s position, you don’t ask 2,200 respondents what they think. You triangulate behavioural evidence from sources Tesla doesn’t control, can’t edit, and didn’t commission: registration data, third-party reviews, used-car residual values, switching behaviour at the dealership, the BYD shipment data, the China revenue trajectory, and the cancellation rate on FSD. That evidence points to a different diagnosis: the position is leaking because the proof is leaking. The CEO’s persona is a small variable in a large equation, and the equation is dominated by what the company is and is not shipping.<\/p>\n\n\n\n

The HBR piece points the operator at the wrong instrument and the wrong layer of intervention. That is the cost of the seven-layer stack going unnamed.<\/p>\n\n\n\n

4 \u2014 The framework: amplifier versus engine<\/h2>\n\n\n\n

Fame and distribution are amplifiers. They take what is already underneath and make it more visible, faster. They are not engines that generate outcomes on their own.<\/p>\n\n\n\n

The distinction matters because the work is different.<\/p>\n\n\n\n

If your noun is paid for, amplification compounds the position. Every additional unit of attention reinforces the noun the customer already associates with you. Justin Welsh writes about solopreneurship, ships solopreneur products, runs his business as a public worked example of the noun, and his audience growth reinforces the noun at every step. The amplification and the proof point in the same direction.<\/p>\n\n\n\n

If your noun is hollow, amplification accelerates the unravelling. Quibi had $1.75 billion in funding, Jeffrey Katzenberg and Meg Whitman at the helm, a $5.6M Super Bowl ad, and peak distribution across every owned and paid channel. The product had no underlying position that the market could memorize. Amplification gave 500,000 people six months to confirm there was nothing to keep watching, and the company died in April 2020.<\/p>\n\n\n\n

Tesla’s amplification has been compounding for fifteen years because the underlying noun was real. The HBR piece treats that amplification as if it were the cause of Tesla’s success, and concludes that the way to fix Tesla is to manage the amplifier. That is the wrong layer of intervention.<\/p>\n\n\n\n

The engine is the position. The position is built by capital allocation and refusal. Musk’s fame neither built it nor can it rebuild it.<\/p>\n\n\n\n

5 \u2014 Why Tesla’s position is actually weakening<\/h2>\n\n\n\n

Tesla’s position is<\/em> weakening. The HBR piece is responding to a real signal. But the cause sits in a place the article never visits.<\/p>\n\n\n\n

The original noun (the future of transportation) is being earned by other companies. BYD is shipping electric vehicles at scale, with vertical integration that rivals Tesla’s, at price points Tesla cannot meet. Hyundai-Kia is shipping vehicles with software stacks that increasingly meet Tesla’s bar. Chinese entrants have closed the gap on Autopilot equivalents. The Cybertruck launched and underperformed against Tesla’s own volume projections. The Model Y refresh did not move the category. FSD continues to be the product Tesla promises but does not ship.<\/p>\n\n\n\n

The proof is leaking. When the proof leaks, the position leaks. When the position leaks, the noun becomes available to whoever ships next.<\/p>\n\n\n\n

That is the diagnosis the financials would agree with. Tesla’s margin compression, the discounting cadence on Model Y inventory, the China revenue trajectory: each one is a leading indicator of position decay through proof decay.<\/p>\n\n\n\n

It has very little to do with how Musk is perceived. It has everything to do with whether the company is still doing the work that earned the noun in the first place.<\/p>\n\n\n\n

6 \u2014 What the HBR piece would have to do to be right<\/h2>\n\n\n\n

For HBR’s thesis to hold, three things would need to be true.<\/p>\n\n\n\n

The first<\/em> is that Tesla’s positioning was authored by Musk’s persona rather than by company decisions. Fifteen years of capital allocation suggest otherwise.<\/p>\n\n\n\n

The second<\/em> is that fame is sufficient on its own to build or destroy a position. Quibi, Fire Phone, Zune, Google+, every celebrity restaurant chain, every founder who ran out the clock on borrowed attention without underlying proof: each is a counter-example. The graveyard is long. It is invisible because failures don’t headline conjoint studies.<\/p>\n\n\n\n

The third<\/em> is that the strategic intervention is at the CEO-perception layer. The piece offers three paths: realign Tesla left of Musk’s politics, double down right, or fix the product. The first two are accommodations of a perceived audience problem. The third is the only one that addresses the real problem, and the article treats it as the least urgent of the three.<\/p>\n\n\n\n

A positioning decision, in irreducible terms, is a refusal. The realignment paths the article offers are not refusals. They are accommodations of stated preference. Tesla did not build its noun by accommodating. It built it by refusing. It will not rebuild it by accommodating now.<\/p>\n\n\n\n

7 \u2014 Finally<\/h2>\n\n\n\n

This is the test I’d hand to any operator reading the HBR piece and feeling its pull:<\/p>\n\n\n\n

Take the company you work on. Strip the CEO from the front of it. Remove every public statement, every interview, every conference appearance, every tweet, every shareholder letter. Now look at what the company has shipped, what it has refused to ship, what it has invested in, what it has divested from, what it has charged for, and what it has given away.<\/p>\n\n\n\n

What noun does the pattern point to?<\/p>\n\n\n\n

If you can name one, the CEO is amplifying a position that exists. The fame is the residue. Managing the persona is a third-order concern.<\/p>\n\n\n\n

If you cannot name one, the CEO is the position. The brand is the costume. Removing the costume removes the company.<\/p>\n\n\n\n

Tesla, with Musk subtracted, still points to a specific noun. The noun is weakening because the proof is weakening. That is the work the company needs to do.<\/p>\n\n\n\n

The article points at the wrong layer.<\/p>\n","protected":false},"excerpt":{"rendered":"

On the difference between the fame that follows proof and the fame that replaces it. The Harvard Business Review just published a piece arguing that Tesla’s brand is now hostage to its CEO, and that the fix is for the company to manage Musk’s perception more carefully. The piece is built on a 2,200-respondent conjoint […]<\/p>\n","protected":false},"author":1,"featured_media":4390,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_coblocks_attr":"","_coblocks_dimensions":"","_coblocks_responsive_height":"","_coblocks_accordion_ie_support":"","footnotes":"","rank_math_title":"","rank_math_description":"","rank_math_canonical_url":"","rank_math_focus_keyword":""},"categories":[76],"tags":[],"class_list":["post-4389","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-positioning"],"_links":{"self":[{"href":"https:\/\/paulsyng.com\/blog\/wp-json\/wp\/v2\/posts\/4389","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/paulsyng.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/paulsyng.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/paulsyng.com\/blog\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/paulsyng.com\/blog\/wp-json\/wp\/v2\/comments?post=4389"}],"version-history":[{"count":1,"href":"https:\/\/paulsyng.com\/blog\/wp-json\/wp\/v2\/posts\/4389\/revisions"}],"predecessor-version":[{"id":4391,"href":"https:\/\/paulsyng.com\/blog\/wp-json\/wp\/v2\/posts\/4389\/revisions\/4391"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/paulsyng.com\/blog\/wp-json\/wp\/v2\/media\/4390"}],"wp:attachment":[{"href":"https:\/\/paulsyng.com\/blog\/wp-json\/wp\/v2\/media?parent=4389"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/paulsyng.com\/blog\/wp-json\/wp\/v2\/categories?post=4389"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/paulsyng.com\/blog\/wp-json\/wp\/v2\/tags?post=4389"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}