Costly Signal

A claim any competitor could make for free carries no information about which one is true. What separates a real position from marketing copy is what it would cost to fake.

September 1, 2026 7 min read

Two competing products both put "the easiest onboarding in the category" on their homepage. A prospect reading both pages learns nothing from either sentence, not because one company is lying, but because both companies could type that exact sentence whether or not it's true. The claim costs nothing to make and nothing to be wrong about, so its presence on a page carries no information about which company actually has the easier onboarding.

A claim that costs nothing to make carries no information

This isn't a complaint about marketing copy being untrustworthy. It's a plainer fact: if a claim is equally cheap for a company with the advantage and a company without it to make, then observing the claim doesn't help anyone tell the two companies apart. Before either page loaded, a prospect's best guess about which company actually has easier onboarding was some number. After reading both claims, that guess should be exactly the same number, because both companies, real advantage or not, would say the identical thing. A signal that any type of sender can produce at the same cost isn't really a signal. It's noise shaped like one.

What would actually move that guess is something the two companies couldn't produce equally easily. Not costly for its own sake, expensive doesn't mean informative on its own, but costly in a specific, lopsided way: cheaper to produce if the underlying claim is true, more expensive or outright impossible to fake if it isn't. A company that genuinely has the easier onboarding can point to deals that closed specifically because of it, prospects who evaluated a harder competitor and switched. A company whose onboarding is not actually easier would have to manufacture that same pattern of deals without the underlying product behind it, and that's the harder, costlier thing to fake convincingly at scale.

Where this gets its name

Economics has a formal name for exactly this asymmetry, developed for a different market entirely. In 1973, Michael Spence published "Job Market Signaling" (Quarterly Journal of Economics), asking how an employer, unable to directly observe a job candidate's productivity, could still learn something real from what a candidate does before being hired. His answer wasn't that costly signals are informative because they're costly. It was narrower and sharper: a signal is informative only when it's cheaper to produce for the type of person the employer actually wants to identify than for the type they don't, so that faking it costs more than it's worth to the person faking it, and only the real thing is worth producing at that price. In his model, education can function this way, not necessarily because it teaches the skills a job needs, but because a genuinely more capable candidate can often clear it at lower cost than a less capable one, holding the reward fixed. Take that condition away, make the cost the same regardless of who's producing the signal, and the same credential stops separating anyone from anyone else.

The honest limit: cost has to be lopsided, not just present

This is the part worth being careful about, because it's easy to round Spence's result down to "make it expensive and it becomes credible," and that's not what the model says. Spending money on a positioning claim, a rebrand, a big campaign, a slick landing page, is a real cost, but if a weaker competitor can spend the same amount to produce the same appearance, the cost bought nothing. What has to be true is that the cost of producing the specific evidence behind a positioning claim runs differently for a company that actually has the advantage than for one that doesn't. A single closed deal citing the claimed advantage isn't enough on its own either, one deal can close for a dozen reasons unrelated to positioning, timing, an existing relationship, a discount. What separates a real signal from a lucky data point is a pattern: many deals, cited for the same reason, that would be expensive and difficult for a company without the underlying advantage to manufacture at the same rate.

Where this fits at Brief

This is the actual shape of the problem Positioning Agent is built to sit on top of: proposing how a company is positioned by drawing on competitors, deals, and signals together, not any one of them alone. A positioning claim checked only against itself is exactly the cheap-talk case, a sentence that costs nothing and proves nothing. Competitor Agent's documented job is narrower than analyzing a rival's claims, it crawls competitor pages and flags when one of them drifts, but that raw material is exactly what a claim about relative positioning would need to be checked against. Pipeline Agent's job is narrower too, it syncs deals from the CRM, but a synced record of real deals is exactly where a pattern like wins clustering around a claimed advantage would have to be found, if it's there at all. This post can't speak to what analysis, if any, gets run on top of that raw material today. What's true regardless of implementation is the shape of the problem: a positioning claim checked only against itself proves nothing, and the only way to check it against anything else is against real, hard-to-fake records like these, not against another sentence.

Next time your team's positioning deck states an advantage, could a weaker competitor produce that exact sentence for the same cost you did, or would it cost them something real to say it and mean it?

Frequently asked questions

What is a costly signal? A claim or action that only makes sense to produce if it's actually true, because producing it while it's false would cost more than it's worth. The cost has to fall differently on the party that has the underlying advantage than on one that doesn't, otherwise the signal is just as cheap to fake as it is to send honestly, and it stops conveying any information.

Where does the term come from? Michael Spence's 1973 paper "Job Market Signaling," which modeled how an employer could learn something real about a job candidate's productivity from actions taken before hiring, when a candidate more capable of the job could often produce those actions at lower cost than a less capable one. The result is general to any market where one side can't directly observe what the other side actually has.

Doesn't spending more on marketing make a positioning claim more credible? Not by itself. A rebrand or a big campaign is a genuine expense, but any competitor with a comparable budget, weaker positioning and all, can buy the identical polish. Nothing about matching spend tells the two apart. What has to differ is the cost of the specific evidence a claim depends on, not the size of the check written to make the claim look good.

Why does Positioning Agent need Competitor Agent and Pipeline Agent instead of working from a claim alone? Because a claim has nothing else to be measured against on its own. What each of the two supplies is narrower than an analysis, real crawled pages from the field a company is claiming to stand apart from, and a real synced ledger of deals that actually closed, but narrow and real is still worth more here than broad and unverifiable. Whatever gets built on top of that raw material, this post doesn't claim to know the specifics, has something concrete to check a stated position against instead of nothing at all.