The marketing that actually earns trust in a skeptical market is the marketing your finance team wants to cut. The 90-day guarantee that exposes you to returns. The genuinely useful free tool you gave away instead of gating. The founder essay that took a week to write and can't be attributed to a campaign. Each looks like waste on a spreadsheet, and each is doing the one job that adjectives and "trusted by" logos cannot. When you run these through an efficiency review and strip the cost out, you don't make the signal cheaper. You delete it.
This is not a branding aesthetic. It's a consequence of a hard result that biology and economics reached independently, and it changes what counts as a good marketing decision.
Why the peacock is not being wasteful
Amotz Zahavi's handicap principle, proposed in 1975 and rejected by most of his field for a decade, made a claim that sounded absurd: animals evolve signals that are credible because they are costly. A peacock's tail is metabolically expensive, mechanically cumbersome, and a liability against predators. That is precisely why a peahen can trust it. A sickly, parasite-laden male cannot grow and haul around a full, symmetric tail. The cost is not a bug in the signal; the cost is the signal. A gazelle that spots a lion often stotts — springs straight up on stiff legs instead of fleeing. It is burning energy and lead time to broadcast a message: I am so fit I can afford to waste this in front of you, so don't bother chasing me. A weak gazelle physically cannot fake the display.
The economists arrived at the same structure. Michael Spence's job-market signaling model — the work that anchored his Nobel — shows why a degree can be worth paying for even if it teaches you nothing relevant to the job. The degree separates candidates only when it is differentially costly: harder, more expensive, more time-consuming for a low-productivity worker to obtain than a high-productivity one. Strip that differential and the signal collapses. If everyone could get the credential at equal cost, employers would learn nothing from it, and it would be rational to stop paying attention.
The formal name for what makes both cases work is a separating equilibrium. A signal transmits information only when high-quality and low-quality senders face different costs to send it. When both can send it equally cheaply, you get a pooling equilibrium: everyone sends the signal, it becomes noise, and rational receivers discount it to zero. That is the whole game. It is also, almost exactly, the state of most marketing.
Most marketing is cheap talk, and buyers know it
"Industry-leading." "Trusted by thousands." "Best-in-class support." "Customers love us." A firm with a genuinely superior product and a firm running a scam can type these sentences at identical cost — which is roughly zero. Game theorists call this cheap talk, and the theorem is unforgiving: when a message costs the same to send regardless of the truth behind it, it carries no information in equilibrium. Your buyer's skepticism toward your adjectives is not irrational cynicism. It is the correct Bayesian response to a pooling signal. They have seen the identical words from your worst competitor, so the words update their beliefs by nothing.
Founders consistently misread this. They assume the problem is that the claim isn't loud enough, isn't repeated enough, isn't phrased persuasively enough, so they spend more to send the same cheap signal at higher volume. Amplifying a zero-information message gives you more zero. The buyer is not failing to hear you. They are correctly refusing to trust a message that a liar could send just as easily.
The signals that move a skeptical buyer are the ones a low-quality seller structurally cannot afford:
- A long, no-questions refund window is nearly free if your product is good and returns are rare. It is ruinous if your product is bad and everyone sends it back. The cost self-selects by quality.
- A genuinely useful free tool requires real engineering and ongoing maintenance. A fly-by-night operator can't sustain the burn; the tool's continued existence is itself evidence of a going concern with something to protect.
- Refusing to discount signals that you expect demand at your price and aren't desperate for cash. The struggling competitor, quietly bleeding, almost always breaks and cuts price. Holding the line is a stott.
- Patient, founder-attributed writing puts a named reputation on every claim over years. It is expensive in the one currency you can't print — the founder's time and standing — and that expense is exactly what makes it legible as sincere.
None of these says "we're high quality." Each is an action that would be irrational for a low-quality firm to take. That is the whole trick. The signal is credible because it would hurt to send if it weren't true.
The market where I watched this run without a safety net
I build Kommerce, an operating system for cash-on-delivery commerce in markets where institutional trust is thin — where courts are slow, chargebacks don't reliably exist, and a buyer handing money to an unknown online seller has almost no recourse if the box turns out to be empty. In the West you can be sloppy about costly signals because the institutions send them for you. Stripe, the card-dispute process, small-claims court, and a functioning returns culture all lower the trust others must extend on faith. Remove those, and you get a clean laboratory for signaling theory, because the buyer has nothing to fall back on but the seller's signals.
In these markets, cash on delivery is not a payment quirk. It is the dominant separating mechanism, and it is a costly signal in the strict Zahavi sense. The seller who offers COD absorbs the return risk, the reverse-shipping cost, and the loss on every refused package. For a seller shipping what they promised, refusals stay manageable — the correlation runs their way. For a seller shipping garbage, refusals compound until the model bankrupts them. COD does not assert honesty; it imposes a cost structure a dishonest seller cannot survive. Buyers there internalized this without any theory. They trust the COD seller and ignore the one demanding prepayment and flashing five-star badges, because they intuit which signal a scammer could fake for free.
The same logic governs which sellers survive on the platform and which quietly die, and it maps onto a point I've argued about defensibility: much of what founders call a moat is really an accumulated stock of costly signals a competitor would have to re-pay from zero, not a magic property of the technology, which is why network effects get overrated as the source of durability. A reputation built on years of honored guarantees is expensive to replicate precisely because the expense was the point.
The test, and the trap
Here is the operational version, usable in a meeting. For any marketing or brand move on the table, ask one question: could a low-quality competitor cheaply fake this?
If yes, it is cheap talk, and you should expect it to signal nothing no matter how much you spend amplifying it. If no — if faking it would cost a bad sender far more than it costs you — it is a real signal, and its apparent inefficiency is load-bearing. When someone proposes "shorten the guarantee to reduce return exposure" or "gate the free tool to capture more leads" or "replace the founder's slow essays with high-velocity AI content," they are proposing to lower the cost of the signal. They will frame it as efficiency. What they are doing is moving you from a separating equilibrium toward a pooling one — toward the noise where your worst competitor already lives.
There is a real boundary here, and skipping it turns the principle into an excuse for burning money. Cost alone does not make a signal. The cost must be correlated with the quality it signals. A handicap works only when it is differentially expensive — cheaper for the good sender, brutal for the bad one. This is why the elegant signals are the ones that are nearly free if you're good and catastrophic if you're not; they self-select. Lighting money on fire in a way a well-funded fraud could match at the same price signals nothing about your quality — it signals only that you have money, which a bad actor may also have. The classic economic defense of ad spend (Nelson; Milgrom and Roberts) survives only under the assumption that the cost is recouped through repeat purchase, which requires the product to actually be good. Break that link and the signal breaks with it. So the test has two clauses, not one: the move must be costly, and the cost must be one only a high-quality sender can comfortably bear.
This is also why importing tactics without the underlying model fails so reliably. Founders copy the visible artifact — the long guarantee, the free tool, the founder newsletter — without preserving the cost structure that made it informative, then wonder why the cargo-cult version converts nothing. It's the same category error I've written about in treating strategy as a set of surface tactics rather than a system with selection pressures: the artifact is downstream of the mechanism, and copying the artifact while optimizing away the mechanism gets you the shape of the thing with none of its function.
So before the next efficiency review touches your marketing, sort every line item by one question — not "what did this cost us" but "what would it cost a liar to send." Protect the expensive answers. They are the only ones your skeptical buyer was ever going to believe.