Advertising as a Costly Signal: Why Expensive Campaigns Persuade Even When They Say Nothing

TL;DR: A large share of advertising persuades not through what it says but through what it visibly cost. Spence's 1973 signaling model gives the mechanism: a costly action separates high types from low types only when the high type's marginal cost of the action is lower, the single-crossing condition. Nelson (1974) and Milgrom and Roberts (1986) carried it into advertising, where the cost is the message: a firm expecting repeat purchases can burn money in public and earn it back, while a firm selling a product buyers will not repurchase cannot. The asymmetry is why "brand" spend that conveys no measurable fact still moves demand, why a Super Bowl spot at roughly $7 million works precisely because everyone knows it cost that much, and why cutting brand budget to fund performance media often raises blended acquisition cost while every performance dashboard reports improving efficiency. The signal is slow, cross-channel, and shows up as a rising baseline rather than an attributed conversion, which is exactly the effect an attribution system is built not to see. The empirical record, from Ackerberg's yogurt panel to Binet and Field's 60:40 finding, supports a signaling channel with sharp limits: it fails in one-shot markets, when spend is unobservable, and when no repeat purchase exists to recoup it.

---

The Ad That Says Nothing

A perfume advertisement runs during a break in the game. Thirty seconds: a famous face, a city at night, a bottle, one word. The spot names no scent, no price, no ingredient, and no reason the fragrance beats the one beside it on the shelf. By every standard a copywriter would recognize, the advertisement carries no information. The placement cost the advertiser somewhere near $7 million before a cent was spent on production, and the campaign works. Sales rise, the brand holds its price, and the marketing team keeps its budget.

The uncomfortable truth is that a large share of advertising works because it says nothing and costs a great deal. The message is not the message. The cost is the message. A firm that spends $7 million to say almost nothing is communicating something a paragraph of product facts cannot: that it expects to be selling this product long enough, and profitably enough, to make $7 million of public expenditure worth the outlay. A fly-by-night operator selling an inferior product would never write that check, because it could never earn the money back. Consumers do not reason this through in words. Markets reason it through for them.

This essay is about the game-theoretic machinery behind that sentence, and about what it implies for the fight over marketing budgets happening inside almost every growth team right now. The formal idea is signaling, and it is one of the most consequential results in economics of the last half-century. Michael Spence won a Nobel Prize for it. The applied version explains why a certain kind of advertising spend that no attribution model can justify is nonetheless the thing keeping the attributable channels efficient, and why the team that cuts it to hit a quarterly cost-per-acquisition target is often destroying the asset that made the target reachable.

Two clarifications before the theory. First, not all advertising is a pure signal, and the essay is careful about which parts are. Some advertising genuinely informs, some builds mental availability in the sense Byron Sharp means, and some does the signaling work described here. These channels coexist, and a real campaign mixes them. Second, signaling is not magic and it is not always operating. The same model that explains when burning money persuades also says precisely when it does not. An operator who takes only "brand spend is a signal, therefore spend more" has learned the dangerous half of the lesson.

What a Signal Has to Cost

Start with the problem signaling solves. A buyer faces a seller who knows something the buyer does not: the true quality of the product. The seller of a good product and the seller of a bad one both say the same words. Talk is cheap, so talk cannot separate them. Akerlof (1970, Quarterly Journal of Economics 84(3), 488-500) showed where that leads: when buyers cannot tell quality apart, they will only pay the average, good sellers exit because the average price does not cover their higher costs, and the market can unravel until only the worst products remain. The market for lemons is the baseline case, the world without a working signal.

A signal is what breaks the unraveling. Spence's insight, from "Job Market Signaling" (1973, Quarterly Journal of Economics 87(3), 355-374), is that a high type can distinguish itself from a low type by taking a costly action, provided that the action costs the two types different amounts. Spence's original setting was education. A worker knows whether they are productive; the employer does not. If a productive worker finds a degree easier or cheaper to obtain than an unproductive one, then a degree can credibly separate them, even if the degree teaches nothing relevant to the job. The content of the credential is beside the point. The differential cost of acquiring it is the whole mechanism.

The condition that makes this work has a name, and it is worth stating precisely because everything downstream depends on it. Let s be the level of the signal and θ the sender's type, with higher θ meaning higher quality. Let c(s,θ) be the cost of producing signal level s for a sender of type θ. Separation is possible when signaling is costly for everyone but less costly at the margin for higher types:

$$\frac{\partial c(s,\theta)}{\partial s} > 0, \qquad \frac{\partial^2 c(s,\theta)}{\partial s\, \partial \theta} < 0.$$

The first inequality says more signal costs more. The second, the single-crossing or Spence-Mirrlees condition, says the marginal cost of an extra unit of signal falls as type rises. When both hold, there is a signal level the high type is willing to reach and the low type is not, and observing that level lets the buyer infer quality. Without the second inequality, signaling collapses: if the two types face the same cost curve, any level one can afford the other can afford too, and the signal separates nothing.

The peacock and the MBA

Biology arrived at the same result independently, and the parallel is not decoration. Zahavi (1975, Journal of Theoretical Biology 53(1), 205-214) proposed the handicap principle to explain the peacock's tail. A large, metabolically expensive, predator-attracting tail seems to make no evolutionary sense until you notice what it proves: only a genuinely fit male can grow one and survive carrying it. The tail is credible as a fitness signal precisely because it is costly, and costly in a way a weak male cannot afford. Grafen (1990, Journal of Theoretical Biology 144(4), 517-546) later showed formally that Zahavi's verbal argument holds as a signaling equilibrium under conditions that are recognizably the single-crossing condition in another notation.

The peacock's tail and the master's degree are the same object. Both are expensive, both are wasteful in the narrow sense that the resources could have gone to something useful, yet both work as signals for exactly that reason. Waste is not a bug in a costly signal. Waste is the feature that makes it impossible to fake. A Super Bowl spot is a corporation's tail.

The applied reader should already feel the shape of the argument. A brand campaign that conveys no product information is not failing to communicate. A brand campaign that conveys no product information is communicating the one thing product information cannot: that the advertiser had the money, and expected the future revenue, to justify spending it in public. The rest of this essay works out when that inference is warranted, when it is not, and what an operator should do about it.

Why Advertising Is a Signal, Not a Message

Spence's model is about people. The move to advertising was made by Phillip Nelson, first in "Information and Consumer Behavior" (1970, Journal of Political Economy 78(2), 311-329) and then decisively in "Advertising as Information" (1974, Journal of Political Economy 82(4), 729-754). Nelson's contribution was to notice that the puzzle of uninformative advertising dissolves once you distinguish two kinds of goods by when the buyer can judge quality.

Search, experience, and credence goods

A search good is one whose quality you can verify before buying: screen size, thread count, horsepower, the number of ports on a laptop. For search goods, advertising works the obvious way. The advertisement lists the verifiable attributes, the buyer checks them, and content carries the information. Signaling has little to do.

An experience good is one whose quality you learn only after using it: the taste of a soft drink, the reliability of a car, whether a film is any good. For experience goods, the advertisement cannot credibly tell you the product is good, because a bad product's maker can print the same words. What the advertisement can do is be expensive in a way you can see. Nelson's argument was that heavy advertising of an experience good is a signal that the firm expects you to become a repeat buyer, because only repeat business makes heavy advertising pay. The content is close to irrelevant. The expenditure is the information.

Darby and Karni (1973, Journal of Law and Economics 16(1), 67-88) added a third category. A credence good is one whose quality you cannot verify even after consuming it: a vitamin supplement, a piece of financial advice, many medical services. For credence goods, neither inspection nor experience resolves the uncertainty, so signals through reputation and sunk cost do even more of the work, and the failure modes are worse.

Nelson's empirical prediction was ordinal and it has held up: experience goods are advertised more intensively than search goods, measured as advertising-to-sales ratio, because the signaling return exists for experience goods and barely exists for search goods. The direction is the claim worth carrying. A category where quality is invisible until purchase is a category where expensive advertising can pay for itself through the repeat business it protects.

Money burning, formally

Nelson's argument was verbal. Kihlstrom and Riordan (1984, Journal of Political Economy 92(3), 427-450), in "Advertising as a Signal," gave it a first formal treatment, showing conditions under which advertising expenditure alone can support a separating equilibrium in which high-quality firms advertise and low-quality firms do not. Milgrom and Roberts (1986, Journal of Political Economy 94(4), 796-821), in "Price and Advertising Signals of Product Quality," produced the version most people mean when they say advertising is a signal. Their key idea is dissipative advertising, or money burning: expenditure that produces no direct benefit to the consumer and exists only to be observed. A firm that lights money on fire in public is credible precisely because the act is pointless for anyone who does not expect future profit.

We can write the mechanism cleanly. Consider a firm choosing an advertising level A. A consumer who observes the spend forms a belief about quality, tries the product if the belief is high enough, and becomes a repeat buyer only if the product turns out to be good. Let m be the margin the firm earns on a purchase, let n(A) be the number of first-time triers a spend of A attracts, and let R be the discounted stream of repeat margin from a satisfied customer. A high-quality firm earns m + R per trier; a low-quality firm earns only m, because its triers discover the truth and never return. A separating equilibrium in which the spend level A* signals high quality requires that the high type profit from it and the low type not:

$$n(A^{*})\,m \;<\; A^{*} \;\le\; n(A^{*})\,\big(m + R\big).$$

The left inequality says the low type cannot recoup A* from one-time sales alone, so it will not spend. The right inequality says the high type can recoup it once repeat business is counted, so it will. The interval between them is where the separating signal lives. The most important feature of the whole essay is hiding in this expression: the interval is nonempty only if R > 0. No repeat business, no signaling equilibrium. The entire mechanism runs on the discounted value of purchases the firm has not made yet.

The repeat value is itself worth writing down, because it is where the discount rate and the churn rate enter. If a satisfied customer repurchases with per-period probability r and the firm discounts the future at factor δ, then

$$R \;=\; \sum_{t=1}^{\infty} (\delta r)^{t}\, m \;=\; \frac{\delta r}{1 - \delta r}\, m, \qquad 0 < \delta r < 1.$$

Read the two equations together and the operating conclusions fall out mechanically. High retention (r near one) makes R large, widens the separating interval, and makes expensive brand advertising rational. Low retention or a one-time product (r near zero) collapses R toward zero, shrinks the interval to nothing, and makes the same spend a pure loss. The math is not decoration on the strategy. The math is the strategy.

A firm that burns money in plain sight is betting it will live long enough to earn it back; a firm selling junk cannot.

, Paraphrasing Milgrom and Roberts (1986)

Milgrom and Roberts also made a subtler point that the practitioner debate usually misses: price and advertising signal jointly. A high-quality firm may signal partly through a high introductory price and partly through advertising, and the mix depends on how much repeat business each protects. A brand that signals only through price leaves demand on the table; a brand that signals only through advertising forgoes margin. The joint problem is why "should we discount or should we advertise" is not two questions but one.

Reading the game tree

The sequence is a signaling game in the formal sense, and the equilibria are Nash equilibria of it, refined to rule out implausible beliefs. Nature sets quality. The firm, knowing its own quality, chooses how much to spend where the market can see it. The consumer, seeing the spend but not the quality, updates a belief and decides whether to try. Quality is then revealed by consumption, and repeat purchase follows or does not.

Loading diagram...

Signaling games are notorious for having many equilibria, including pooling equilibria in which both types choose the same spend and the signal conveys nothing. Cho and Kreps (1987, Quarterly Journal of Economics 102(2), 179-221) supplied the standard tool for discarding the implausible ones, the intuitive criterion, which asks whether a deviation could only have come from a type that would benefit from it and prunes beliefs accordingly. Under that refinement the separating equilibrium, where high quality spends and low quality does not, is the one that survives in the cases we care about. The practical residue: a market can get stuck pooling, where nobody's advertising means anything, and the way out is for a high-quality firm to spend past the level a low-quality firm could ever justify.

The Evidence, in Order

Signaling is an unusually clean theory, which makes it unusually easy to over-apply. The empirical literature is where its real scope gets settled, and in my reading it is more supportive than skeptics expect yet more bounded than enthusiasts want.

The review that set the terms

Kirmani and Rao (2000, Journal of Marketing 64(2), 66-79), in "No Pain, No Gain," organized two decades of work on signaling unobservable quality into a framework practitioners can actually use. Their central distinction is between signals that put money at risk regardless of outcome and signals whose cost is paid only if the firm fails to deliver. A money-burning advertising campaign is the first kind: the cash is gone whether or not the product is good, and its credibility comes from the fact that a low-quality firm could not afford the sunk loss. A warranty is the second kind: it costs the firm nothing if the product performs and a great deal if it does not, so it is credible for a different reason. The review's value is in showing that "signal" is not one lever but a family, and that advertising sits specifically in the revenue-risking, sunk-cost corner. An operator who treats a warranty and a brand campaign as interchangeable trust-builders has missed which risk each one is pledging.

Brand equity as a signal

Erdem and Swait (1998, Journal of Consumer Psychology 7(2), 131-157) reframed the marketer's central asset, brand equity, as a signaling phenomenon rather than a bundle of associations. In their account a brand is a credible summary of everything the firm has done, and its value to the consumer is that it lowers both perceived risk and information cost. The dimension that matters most in their framework is credibility, which they decompose into trustworthiness and expertise, and credibility is built by consistent past investment that would be irrational to have made without a good product behind it. Brand equity, on this reading, is accumulated signal. Every past campaign and every consistent delivery is a deposit, and the balance is what lets a new product from the same brand be believed before anyone has tried it. That is why a brand extension can launch on a fraction of the advertising a new entrant would need: it is drawing down a signal already paid for.

The yogurt panel

The most direct test of the informative-signaling channel against its main rival, the pure prestige or image effect, comes from Ackerberg (2001, RAND Journal of Economics 32(2), 316-333), "Empirically Distinguishing Informative and Prestige Effects of Advertising." Ackerberg used household scanner-panel data around the launch of a newly introduced yogurt and split consumers by whether they had already tried the brand. The logic of the test is clean. If advertising works by informing or by signaling to the uninformed, it should move the behavior of consumers who have not yet tried the product and do little for those who already know it firsthand. If advertising works by conferring prestige or social image, it should keep affecting experienced users too, because the image is consumed on every purchase.

The data came down on the informative side. Advertising raised the purchase probability of consumers who had never tried the brand and had no measurable effect on those who had already experienced it. For an experience good discovered through trial, the advertisement did its work by getting the uninformed to try, exactly what a signal is supposed to do, and stopped mattering once experience took over as the better source of information.

What the survey concludes

Bagwell's survey, "The Economic Analysis of Advertising" (2007, in Handbook of Industrial Organization, Vol. 3, 1701-1844), is the authoritative map of this territory. Its verdict on the signaling literature is measured: the money-burning mechanism is theoretically sound and empirically relevant, but it is one of several channels through which advertising operates, it is strongest for experience goods with meaningful repeat purchase, and its predictions about the relationship between advertising and quality are more fragile than the simplest models suggest. A responsible reading of the evidence is not "advertising is a signal" but "for experience goods sold to repeat buyers, a substantial part of what advertising does is signal, and that part is invisible to any measurement that looks only at the content or the click."

Money burning versus information design

Signaling by cost is not the only way a sender can move a receiver's beliefs. Kamenica and Gentzkow (2011, American Economic Review 101(6), 2590-2615) formalized the other way, in which the sender designs an informative disclosure rather than an expensive gesture, and the content of the message does the persuading. The distinction is worth holding precisely because a real marketing budget spans both.

The two mechanisms make opposite demands. A money-burning signal must be expensive and observable and can be content-free. A persuasion-by-information disclosure must be informative and credible and can be cheap. A ratings badge, a published benchmark, a security audit, or a transparent methodology page is persuasion by information; a marquee sponsorship is persuasion by cost. Confusing them is a common and expensive error. A firm that answers a credibility problem with more disclosure when the market wanted a costly commitment, or with an expensive campaign when the market wanted verifiable facts, spends into the wrong channel and wonders why the needle does not move.

The Money-Burning Machines

The clearest real-world instances of dissipative advertising are the placements everyone knows are expensive. A Super Bowl spot is the canonical case, and its power is inseparable from the public knowledge of its price. When the audience knows a message cost roughly $7 million to air, the price is part of what is received. A cheaper placement conveying the identical creative would signal less, because it would prove less about the advertiser's expected future.

The trade press reports the going rate for a 30-second spot each year, and the trajectory is its own small monument to signaling inflation. The figures below are approximate and drawn from advertising trade reporting; I treat them as orders of magnitude, not audited prices.

The price roughly doubled over the period, yet the audience did not, which tells you the scarcity being sold is not attention alone but the commonly observed fact of the expense. Veblen (1899, The Theory of the Leisure Class) named the individual version of this a century earlier: conspicuous consumption, spending whose purpose is to be seen as spending. A Super Bowl buy is conspicuous consumption by a corporation, and the audience it is really addressing includes retailers, investors, employees, and rivals as much as it does the household on the couch. A brand that can afford the most expensive thirty seconds in media is telling its distribution partners it will still be here next year.

Marquee placement generalizes the point. A billboard in Times Square, a stadium naming right, a jersey sponsorship, a full-page in a newspaper that still commands one: each is chosen partly because the price is legible to the people who see it. The waste is doing the work. An operator evaluating such a buy on cost-per-thousand-impressions against a programmatic alternative has priced the impressions and missed the signal, which is the larger part of what is being bought.

When the Signal Fails

The theory earns its keep by telling us exactly when burning money persuades nobody. Each failure mode violates one of the assumptions the separating interval quietly relies on, and each maps onto a business situation we can recognize in the field.

Four ways the signal breaks

The first failure is equal cost across types. Single-crossing is the whole engine; if a low-quality firm can advertise as cheaply and profitably as a high-quality one, the spend separates nothing and the market pools. Cheap, abundant, untargeted inventory is prone to this. When anyone can buy the same reach for the same price and recoup it from a single sale, the reach stops proving anything about quality.

The second failure is unobservable spend. A signal must be seen to signal. Advertising that reaches each person privately, that no third party observes, and that the recipient cannot tell is expensive carries no cost information. Much precisely-targeted performance media has this property: the individual served an ad has no idea whether it cost the advertiser two cents or two dollars, so the spend, however large in aggregate, transmits nothing about the firm's confidence. Signaling needs an audience that knows others are watching the same expensive thing.

The third failure is the one-shot market. Return to the separating interval: it is empty when R = 0. A market with no repeat purchase gives the high-quality firm no future to recoup the burn, so it will not burn, and the signal never forms. Products bought once in a lifetime, categories driven by one-off events, and businesses whose customers exit after a single transaction all sit here. In a one-shot market the honest advice is often that brand advertising cannot pay, and performance media that harvests existing intent is the correct instrument.

The fourth failure is the operator's version of the third: direct-to-consumer businesses with no genuine repeat purchase. A brand can look like an experience good and behave like a one-shot good whenever churn is high enough that the discounted repeat value R never accumulates. The peacock's tail is a bad investment for a mayfly.

Brand, Performance, and the Signal You Can Break

Now the fight that started this essay. Inside almost every growth team I have advised there is a standing argument between brand and performance, and it is usually conducted as a measurement dispute: performance media is accountable and brand is not, so, the argument goes, budget should flow to what can be measured. Signaling theory says this framing quietly assumes the conclusion, because the thing performance measurement is best at seeing is exactly the thing brand spend is worst at producing, and the thing brand spend produces is exactly what performance measurement is built to ignore.

The budget fight

The practitioner evidence most often cited here is Binet and Field's IPA study, The Long and the Short of It (2013), which analyzed the Institute of Practitioners in Advertising effectiveness databank and reached a now-famous split: across many campaigns, the allocation that maximized long-run business effects was on the order of 60 percent to brand building and 40 percent to sales activation. Brand building, in their data, produces effects that build slowly and persist for years; activation produces sharp spikes that decay within weeks to months. A plan weighted heavily toward activation looks efficient quarter to quarter and underperforms over the cycle.

The 60:40 finding is practitioner evidence and should be read as such, with its limits stated plainly. The databank is a sample of campaigns entered for effectiveness awards, which is not a random sample of all marketing; the optimal ratio varies by category, by growth objective, and by how much brand equity a firm already holds; and the split is an average across conditions rather than a law. A subscription business with strong retention and a new entrant in a one-shot category should not run the same ratio, and the signaling math above says exactly why. What survives the caveats is the direction and the mechanism: for repeat-purchase categories, the channel that pays over years is systematically the one that pays worst on a last-touch report.

What the dashboard cannot see

Here is the mechanism that makes the budget fight dangerous rather than merely contested. In practice, it is the part teams discover too late, after the reallocation is already done. Signal effects are slow, cross-channel, and show up as a lift in the baseline, the demand that arrives without a traceable touch, rather than as an attributed conversion. A performance channel's measured efficiency is computed on the conversions it can claim, and a large share of those conversions are people the signal already moved, who then arrived through a search or a retargeting ad that took the credit. Cut the signal and two things happen on different clocks. The performance dashboard keeps reporting the same or better efficiency for a while, because it is still harvesting the demand the old signal created. The baseline erodes underneath, slowly, and blended acquisition cost, the number that actually matters, rises even as every attributed channel looks fine.

The allocation question this raises, how to size brand against performance as a portfolio with different risk and time profiles, is a problem of its own, and the portfolio-theoretic treatment lives in a separate piece on brand-versus-performance budget allocation. The point here is narrower and prior to allocation: before you can weigh the two, you have to measure the one the attribution stack is structurally blind to, and that requires incrementality and geo-lift testing and marketing-mix modeling rather than last-touch reports. Measure the signal with the wrong instrument and you will conclude, wrongly, that it does nothing.

What This Changes for an Operator

The theory is only worth the decisions it changes. Here are the ones I would change.

Classify the good before you argue about the budget. The single most useful question in the brand-versus-performance debate is not "what is our CAC" but "does our product have real repeat purchase, and can prospects verify quality before buying." A search good with one-time purchase should skew hard to information and harvest; an experience good with strong retention has a genuine signaling return and should protect its brand spend. The separating interval is empty or wide depending on R, and R is a fact about your product, not a preference about your marketing.

Spend where the cost is observable, or do not call it a signal. If the goal of a budget line is to signal confidence, the placement has to be one whose expense the relevant audience can see. Precisely-targeted, privately-served media can be excellent at harvesting demand and is nearly useless as a signal, because nobody watching knows what it cost. Match the instrument to the intent.

Measure signal effects as baseline lift, not as attributed conversions, and give them time. The right instruments are geo holdouts, marketing-mix models, and long-horizon incrementality tests; the wrong instrument is any last-touch or even multi-touch attribution report, which by construction credits the harvesting channel with the signal's work. Judge brand spend on movement in the demand that arrives untouched, over quarters, not on a conversion path over days.

Watch blended cost, not channel cost, when you reallocate. The failure this essay warns about is invisible in the numbers most teams watch. Track total acquisition cost across all channels against total new customers, and treat a divergence between improving channel efficiency and worsening blended cost as the alarm it is: the sign of a signal being spent down.

Hold the line in the argument. When a dashboard-driven case is made to cut brand spend to fund performance, the counter is not a defense of brand as an act of faith. The counter is the mechanism: performance largely harvests demand, the signal largely creates it, and the attribution system that makes performance look efficient is the same system that cannot see the demand creation it is quietly consuming. Such an argument rests on evidence, and it is the one worth having.

None of this is a license to spend. Signaling is a specific mechanism with specific preconditions. The operator's job is not to believe in brand or to believe in performance; it is to know which game the product is in, and to spend into the mechanism that game actually rewards.

Key Takeaways

1. Much advertising persuades through its visible cost rather than its content. Spence's single-crossing condition (1973) is the reason: a costly action separates high from low quality only when the high type's marginal cost is lower, which for advertising means only a firm expecting repeat purchases can recoup a public burn.
2. The separating interval n(A*) m < A* ≤ n(A*)(m+R) is nonempty only when the discounted repeat value R is positive. No repeat purchase, no signaling equilibrium, which is why the whole mechanism is a fact about your product's retention, not about your creative.
3. Nelson (1974) predicted, and the record supports, that experience goods are advertised more intensively than search goods, because signaling pays only where quality is invisible before purchase and durable after it.
4. Ackerberg (2001) found advertising for a new yogurt moved consumers who had never tried it and not those who had, evidence for an informative-signaling channel over a pure prestige effect.
5. Binet and Field's 60:40 brand-to-activation finding (2013) is practitioner evidence that the channel paying best over years reports worst on last-touch attribution; the ratio is an average with real limits, not a universal law.
6. The signal fails in four recognizable cases: costs equal across quality, spend the audience cannot observe, one-shot markets, and high-churn direct-to-consumer businesses where repeat value never accumulates.
7. Cutting brand spend to fund performance can raise blended acquisition cost while every performance dashboard reports improving efficiency, because performance harvests demand the signal created; measure the signal as baseline lift with geo holdouts and marketing-mix models, never with attribution.

Further Reading

• Brand vs. Performance: A Portfolio Optimization Framework, once you accept the signal is real, sizing it against performance is a portfolio problem with distinct risk and time profiles; this is the allocation math.
• Category Entry Points and Mental Availability, the demand-side complement to signaling: whom the signal has to reach, and when, for it to be remembered at the moment of purchase.
• Trust Signals and Their Measurable Lift, the micro, on-page counterpart of the macro signal, where the currency is verifiable cues rather than conspicuous cost.
• The Hidden Cost of Optimization and Brand Equity, what happens to the signal when an over-fitted acquisition engine spends the accumulated equity down.
• The Prisoner's Dilemma of Discounting and the Promotion Trap, the companion game, where price rather than advertising is the move, and short-term incentives destroy long-term value.