When one side knows more

01 · THE INFORMATION GAPWhat the other side knows that you don't

Most lessons assume both players know the payoffs — you see the full matrix. Reality is messier. In many of your most consequential decisions, the other side holds private information you can't directly observe.

A patient can see your waiting room, your tone, your website — but not your actual surgical skill. A potential SaaS buyer knows exactly how much time they'd save with Pro features, but you don't. A job candidate knows whether they'll show up and work hard; you're guessing from a 45-minute call.

Adverse selection happens when you can't distinguish types, so you price or treat them as average — but average terms drive away the good types and attract the bad ones. George Akerlof showed this formally with used cars: if buyers can't tell a lemon from a gem, they'll only pay average price, so owners of gems pull their cars from the market, and the market fills with lemons.

Your version: if surgeons can't credibly signal their actual skill, patients default to price or proximity. The high-quality surgeon who charges a premium gets crowded out by the cheaper mediocre one. The signal problem isn't academic — it's eating your referrals.

Two tools break this deadlock, depending on which side of the information gap you're on:

• Signaling — the informed side takes a costly action to credibly reveal its type.
• Screening — the uninformed side designs a menu that causes the other side to self-select and reveal its type.

Adverse selection and information asymmetry introduced formally in SEP — Game Theory; the lemons model and signaling explored throughout Open Yale ECON 159 (Polak).

02 · SIGNALINGThe informed side proves itself

You are the informed party — you know your own surgical skill, your product quality, your commitment. The other side doesn't. Words alone can't solve this, because anyone can say "I'm the best." A signal only works if it's costly in a way that fakers can't afford.

The foundational idea — Michael Spence's single-crossing condition (Job Market Signaling, 1973): a signal credibly separates types if and only if it costs MORE for the low type than the high type to produce it. The high type can afford to signal; the low type can't rationally imitate.

Signals you already hold:

• Board certification & fellowship training. Genuinely expensive to earn for the competent (years, exams, reputation risk). Far more expensive — perhaps impossible — for the incompetent to fake. This is exactly why it works as a credential signal.
• Published skull-base outcomes. Putting your GTR rates, KPS scores, and complication data in public talks or papers is costly if your numbers are bad. A surgeon with poor outcomes simply can't afford to publish them. So publishing credibly signals quality.
• A generous, no-questions money-back guarantee on your SaaS. For a genuinely good product with low churn, offering a 30-day refund is nearly free — almost nobody takes it. For a bad product, a refund guarantee is financial suicide. So only confident sellers offer it.
• Free trial on your best features (briefly). Letting users experience the real value before paying is costly for a product with nothing to show — and cheap for one that delivers immediately.

Notice the pattern: the signal works precisely because it's asymmetrically costly. The good type pays a manageable price to be believed; the bad type pays an unbearable price to imitate. That asymmetry is the whole mechanism.

Spence's model: Michael Spence, Job Market Signaling (1973); signaling equilibria formalized in SEP — Game Theory.

03 · IS IT A CREDIBLE SIGNAL?Apply the test

04 · SCREENINGThe uninformed side designs the choice

Now flip the table. You're the uninformed party — you don't know each buyer's true willingness-to-pay, each patient's seriousness, each hire's effort level. You can't ask directly (people lie or don't know), but you can design a menu of choices that makes each type self-select into the option that fits them.

This is screening. The famous economic archetype: an insurance company offers policies with different deductibles. High-risk customers prefer low deductibles (they expect to claim). Low-risk customers accept high deductibles for lower premiums. By offering both options, the insurer learns risk type without asking — each type reveals itself through choice.

Think about what your three tiers actually do:

• Free tier — a user who just wants to try the product self-selects here. You learn: low commitment or early in the funnel. You lose almost nothing serving them and gain a name in your pipeline.
• Pro ($29) — a doctor who uses the product weekly and sees clear time savings self-selects here. You learn: willingness-to-pay is moderate, usage is real. You're capturing surplus you'd have left on the table with a single price.
• Expert ($59) — a power user running a busy practice or teaching department self-selects here. You learn: high willingness-to-pay, high volume. They're telling you exactly who they are by clicking Expert — you never had to ask.

The screen works because each type makes the choice that's optimal for them — and in doing so reveals private information to you. The menu is designed so that a power user wouldn't choose Free (too feature-limited) and a casual user wouldn't choose Expert (not worth it). Self-selection does the sorting.

Same logic, other contexts:

• Hardcover then paperback: readers who want the book now reveal high willingness-to-pay; patient readers reveal low.
• Early-bird course pricing: physicians who register early reveal higher commitment (fewer last-minute dropouts).
• Deductibles in health insurance: patients who choose lower deductibles reveal higher expected usage.

Screening and mechanism design in SEP — Game Theory; self-selection menus in Open Yale ECON 159 (Polak); tiered pricing as screening — Dixit & Nalebuff, The Art of Strategy (Norton, 2008).

05 · DESIGN THE SCREENWhy the Free tier exists

06 · KNOW THE LIMITSWhen a signal or screen fails

Both tools break under specific, predictable conditions. Knowing the failure mode is what stops you from trusting a signal that's actually noise, or a menu that's actually blind.

Before you lean on either tool, ask which failure mode applies: is the "costly" action actually costly for the type you're trying to screen out, and are your types different enough to sort at all?

07 · LOCK IT INSignaling & screening — the two-move toolkit

• Hidden information is the problem — one side knows something the other doesn't, and naive markets fail (adverse selection).
• If you're the informed side: signal with a costly, hard-to-fake action. The signal works only because low types can't afford to imitate it — this is the single-crossing condition.
• If you're the uninformed side: screen with a menu. Design it so each type finds a different option optimal, and they sort themselves — revealing their type through self-selection.
• Words alone ("trust me", "best quality") are cheap talk — no cost asymmetry, no signal.
• Real signals: board certification, published outcomes, money-back guarantees, warranties, free trials on genuinely good products.
• Real screens: SaaS tiers, insurance deductibles, hardback-then-paperback, early-bird pricing.

Primary sources: SEP — Game Theory (information asymmetry & signaling) · Open Yale ECON 159 (Polak). Named work: Michael Spence, Job Market Signaling (1973) · Dixit & Nalebuff, The Art of Strategy (Norton, 2008).