IGNITE EXTENDED / LESSON 12
What explains a market outcome beyond the observed price?
A market-mechanism sketch and a causal-test note.
A buyer’s reservation value is the most the exchange is worth to that buyer under the stated conditions. The seller’s relevant cost includes the value of the next best use of the resource. If the buyer values the unit more highly, exchange can create surplus.
The transaction price determines how the parties divide that surplus. A price that one party rejects can prevent a mutually beneficial trade. The classroom market exercise also shows why information, timing and rules influence whether potential surplus becomes an actual exchange.
A participant may know their own willingness to pay or cost while remaining uncertain about the other side. Revealing the true reservation value can affect bargaining, so statements about value reflect incentives as well as private information.
A market mechanism determines how offers become trades and what participants can learn from observed behavior. A failure to trade does not necessarily mean that no beneficial exchange existed. Search costs, uncertainty about quality or a strategic attempt to obtain a better price may explain it.
Illustrative supply and demand: quantities become attractive to sellers and buyers at different prices.
A statement can be easy to make regardless of quality. A signal is informative when taking the action has different consequences for different underlying types.
An interested party says a product is reliable.
An action or commitment may provide additional information, but its meaning depends on incentives and the cost of imitating it.
Illustrative application: “our service is reliable” is a claim any seller can make. A meaningful service commitment might expose an unreliable provider to higher expected cost. Whether it informs the buyer depends on the terms, enforceability and incentives, not the mere presence of a guarantee label.
Ask why a low-quality provider would find the action harder or less attractive. Otherwise the supposed signal may carry little information.
The standard diagram: a downward-sloping demand curve and an upward-sloping supply curve cross at equilibrium price P* and quantity Q*.
The crossing assumes buyers and sellers know enough to find each other and trade at that price. The rest of the lesson asks what happens when information, incentives or rules stop the market from reaching it.
The setup: producers, consumers and traders in several separate markets. A producer's widget is worth its production cost if unsold. A consumer earns a private reservation value for one widget only. Players who end in debt must pay.
Each participant knows only their own number. Every gain from trade exists on paper before play starts; whether it is realized depends on how much people reveal through bids and asks.
Eight rules: bids from $0 to $10 in $0.10 steps; the system shows the highest bid and lowest ask; server queue order can override a higher bid; trades happen by taking a standing bid or ask; all bids and asks clear after each trade; a clock runs; balances cannot fall below −$20.
The rules are the market mechanism. Clearing all offers after each trade and showing only the best bid and ask limit what players learn. Changing those rules would change which trades happen, even with identical values and costs.
Step-like supply (rising from about 0.6 to 1.9) and demand (falling from about 2.0 to 0.9) across units 1–6, crossing between units 3 and 4 at about 1.3–1.4.
Built from players' private costs and values, the curves predict roughly three to four trades near the crossing price. Compare that with what actually traded: missed trades and outlier prices show the effect of hidden information and strategic play.
Five obstacles: lack of information (above the surface); no incentive to reveal private information or to trust others (below the surface); strategic behavior, indirect signaling and posturing (subliminal); value perceptions and expectations; agency problems (who decides, and with what incentives).
Use it as a diagnostic checklist when a deal that should create value does not close. Agency problems matter for B2B sales: the person using the product is often not the person paying for it.
Suppose customers who receive a sales call purchase more often. The call may cause purchases, or salespeople may select customers already likely to buy. A third factor, such as urgency, may influence both the call and the purchase.
A causal question asks what would happen to the same relevant population under a different action. Random assignment can make treatment and comparison groups similar on average before the action. The interpretation still depends on implementation, sample size and the outcome being measured.
A statistically or practically noticeable difference does not explain why it occurred. A change in a message could attract a different type of customer, alter expectations or shift purchases in time. The immediate metric may omit a later consequence.
Choose an outcome connected to the business decision and check relevant downstream effects. Repeat purchasing, service burden and refunds can matter when a treatment increases initial conversion. The economics source encourages using experiments to understand behavior rather than treating a single winning variant as the end of analysis.
An observed association does not identify which change produced the outcome. The business decision usually requires a causal explanation.
Customers receiving a tutorial retain better.
Did the tutorial improve retention, or were more motivated customers more likely to take it?
Illustrative application: trained salespeople close more deals. Training may help, but experienced or motivated people may be more likely to take it. Before budgeting for training, investigate a comparison that separates its effect from pre-existing differences and changes in customer mix.
Design the test around the intervention the organization can choose, the outcome that matters and plausible alternative explanations.
Start with a causal explanation and choose an intervention that can challenge it. A result is easier to interpret when the comparison is planned in advance.
State the mechanism: customers abandon signup because the permission request appears before they understand the benefit. Specify the target population and the behavior the explanation concerns.
Compare the current sequence with a revised sequence while keeping other material differences controlled. Use a suitable assignment process so that a pre-existing difference between groups is less likely to explain the result.
Measure completion and the consequences that matter, such as later activation, support requests and errors. A higher click rate can coexist with poorer downstream customer outcomes.
Check whether the result is consistent with the proposed mechanism and whether other explanations remain. Record the sample, conditions and uncertainty before generalizing beyond the test.
A measured difference supports a narrower conclusion than “the new version is better everywhere.” Scope and implementation affect what the experiment can establish.
A discussion slide with two questions: what information was missing, and were there any indirect signals?
Two diagnostic questions to ask after any failed exchange. Indirect signals such as the timing of a bid or a refusal to move on price carry information only when faking them would be costly.
Text: design tests that capture behavior and record enough dimensions to trace causation; A/B testing is essential, but invest in learning why people prefer A to B.
An A/B result tells you which variant won under the test's conditions. Without the mechanism you cannot predict whether it will hold for another segment, season or channel, which is usually the decision you face.
Selected reading: PDF pages 1, 2
The handout specifies a simplified market in which people trade identical widgets while holding different private costs and reservation values. Its rules make incentives and information visible for analysis.
A producer begins with one widget and a production cost; a consumer has a reservation value; a trader can buy and resell.
A transaction occurs when an offer is accepted or compatible prices meet. After a trade, participants must submit fresh bids and asks.
Participants do not know all others’ costs, values or exact numbers. Their offers and observed transactions provide information during play.
Production is fixed at the initial endowment and a participant must own a widget before selling it. These rules shape the possible outcomes.
Selected reading: PDF pages 1, 2
The handout defines the game. Claims about price convergence, efficiency or actual participant behavior require observations from a run.
A reseller’s earlier purchase price is relevant to accounting but a resale choice also depends on the current alternative use or value of the widget.
Identical goods and fixed production remove important real-world complications. State which mechanism the exercise isolates before transferring its conclusion.
The archived exercise gives participants different values, costs and trading roles. Visible bids and asks coexist with private information. The resulting trades can be examined through supply, demand, incentives and beliefs.
Which constraints explain a missed trade?
The FTC's rule on consumer reviews and testimonials took effect on 21 October 2024. It bars buying, selling or disseminating reviews that misrepresent the reviewer's existence or experience, paying for reviews that express a particular sentiment, and suppressing negative reviews. Civil penalties were up to $51,744 per violation when the rule took effect and are adjusted annually for inflation. A testimonial-based signal must therefore be genuine as well as persuasive. Application depends on current law and facts.
A 2025 PNAS study by Sean Westwood (Dartmouth) built an AI agent that completed online surveys with a consistent persona. It evaded 99.8% of standard detection checks at about five cents per response, and could be steered to confirm a hypothesis. Customer-discovery surveys run on open online panels need verified samples, behavioral data or in-person checks before they count as evidence.
Sources: Morgan Lewis: FTC issues final rule on consumer reviews and testimonials · Dartmouth: Sean Westwood wins PNAS prize for AI polling study · 404 Media: A researcher made an AI that completely breaks the online surveys scientists rely on
Fictional market: buyers value a unit at $9, $7 and $4; sellers’ costs are $2, $5 and $8. Each can trade one unit. Assume no transaction cost and full information for the first pass.
Identify mutually beneficial matches. Then add a reliability uncertainty and explain what additional information would matter.
Relay's pitch says it refills about 40% of a clinic's ~42 weekly empty slots, worth ~$96,000 a year gross. A skeptical multi-location buyer asks: how many of those slots would have been filled anyway? The front desk phone tree already fills 10–15%. The three pilot clinics were chosen because their managers were enthusiastic. Fill rates are seasonal. Front desk staff also have their own incentives: some worry that Relay replaces them; others may log phone fills as Relay fills, or the reverse.
Write a one-page causal-test note. Include the counterfactual, the unit of randomization, the primary metric (net incremental visits), the guardrail metrics, how front desk incentives will be aligned, and the result that would make you change the sales pitch.
A fill-rate dashboard credits Relay with every slot it fills, including slots the phone tree would have filled and patients moved up from a later booking. The decision needs the difference from the counterfactual: completed visits that would not have happened otherwise.
Pilot and early-adopter clinics differ from the rest: more motivated managers, deeper demand, better data. Before-and-after or adopter-versus-non-adopter comparisons mix Relay's effect with those differences. Randomize at the cancellation or clinic-day level instead.
When front desk staff decide which channel fills a slot and also log the result, their incentives shape the data. Capture fill source automatically and reward total filled visits so the measurement does not depend on someone's job worry.
A Relay causal-test note: cancellation-level randomization against the phone tree, net incremental visits as the primary metric (≈ $66k/yr illustrative vs the $96k headline), guardrails, and an incentive plan for front desk staff.
Next lesson: Game theory: commitments, trust & coordination
Check your understanding
Question 1
Buyers value one unit each at $9, $7 and $4; sellers' costs are $2, $5 and $8. With full information and no transaction costs, what is the maximum total surplus?
Efficient trades pair the highest values with the lowest costs: (9 − 2) + (7 − 5) = 7 + 2 = $9. Any pairing of those two buyers with those two sellers gives the same total. Adding the $4 buyer with the $8 seller would subtract $4, which is how $5 arises.
Question 2
Relay's dashboard credits it with 17 fills/week at a clinic with 42 empty slots. Before Relay, the phone tree filled 12.5%. At $110 per visit, roughly what weekly revenue can Relay credibly claim as incremental?
The phone tree would have filled 42 × 0.125 = 5.25 slots anyway. Incremental fills ≈ 17 − 5.25 = 11.75; using the 40% rate gives 11.55. Either way ≈ $1,270/week. $1,870 counts every fill as caused by Relay; $578 is the baseline value.
Question 3
Clinics that adopted Relay early show 30% higher revenue per therapist than clinics that did not. An investor calls this proof of impact. What is the strongest objection?
Adopters chose themselves. Strong managers and deep demand drive both adoption and revenue. Correlation can still be useful for prediction, so the last option overreaches. The fix is a comparison in which assignment does not depend on clinic quality.
Question 4
A clinic owner doubts Relay's 40% fill claim. Which offer would be the most informative signal of Relay's quality?
A signal works when it is costly for a low-quality seller to copy. Performance-contingent pricing loses money for a vendor whose product does not work. Claims and testimonials cost little to produce, whatever the true quality. Marketing spend says more about funding than about fill rates.
Question 5
An A/B test shows a new reminder text cuts no-shows by 15% in January at three Austin clinics. The team wants to roll it out nationally. Which concern is most important before scaling?
Randomization identifies the effect under the tested conditions. It does not explain why the effect occurred or guarantee it carries over to other seasons, regions or patient populations. Statistical significance tests chance, not whether the result generalizes. Measuring the mechanism (for example, whether patients replied to reschedule) helps predict where it will hold.
Question 6
A multi-location PT group's operations VP loves Relay, but the deal stalls. The front desk leads, who would run it daily, are lukewarm. Which obstacle from the lesson best describes this?
The surplus belongs to the group's owners, while the effort and perceived risk fall on front desk staff, who can quietly block adoption. Fixing it means changing their incentives and workload, not cutting price or adding information.
Workbook
Module output: A business-model map, market/game models and an advantage-and-pivot memo.
Sources
Original pages and full documents are in the Library edition.