Skip to content

Smoke Test

Probe demand for a product that doesn't exist yet with a cheap false-front — a landing page, pre-order, or fake door — that extracts a commitment-bearing signal, and build only if that signal clears a kill threshold set in advance.

Core Idea

A smoke test, in the lean-startup sense, is a low-cost probe of demand for a product that does not yet exist — a landing page, a pre-order form, an advertisement with a sign-up button, a Wizard-of-Oz mock — that collects a commitment-bearing signal (a click, an email address, a card pre-authorisation, a deposit) approximating real demand without requiring the costly construction of the actual product. The structural commitment is cheap-signal-before-expensive-build: the team designs the cheapest test that still produces a signal commitment-bearing enough to discriminate between real and notional demand, uses the result to decide whether to build, and discards the candidate if the signal falls below the kill threshold. The crux is the gradient between a cheap, unreliable signal — a survey asking "would you buy this?" — and an expensive, reliable signal — the real product in the hands of paying customers — and the smoke test is placed deliberately on that gradient at the point where the cost of the test is low enough to run quickly but the commitment demanded of the respondent is high enough to be predictive. The negative result is as informative as the positive: a landing page that attracts no sign-ups, a pre-order form that collects no deposits, or a Wizard-of-Oz whose manual fulfilment draws no repeat requests all constitute valid kills that save the cost of the full build. The term has a separate, older meaning in software engineering — a quick check that a new build runs at all before deeper testing is committed — which shares the cheap-fast-filter-before-expensive-work structure but tests function rather than demand.

Structural Signature

Sig role-phrases:

  • the expensive commitment — the costly product build or launch the test is deciding whether to undertake
  • the cheap probe — the low-cost false-front artifact that collects the signal (landing page, fake-door, pre-order form, Wizard-of-Oz mock) without building the real product
  • the commitment-bearing signal — the respondent action whose cost approximates real demand (a click < an email < a pre-authorised card < a deposit), positioned on the commitment gradient
  • the cost-vs-predictiveness placement — the probe deliberately set on the gradient where the test is cheap enough to run now yet the demanded action is costly enough to discriminate real from notional demand
  • the falsifiability requirement — the probe must be capable of failing; one that can only pass (a costless button, a leading question) is uninformative and must be redesigned
  • the signal-to-demand model — the (often implicit) conversion mapping from probe response to predicted real demand, only as trustworthy as the probe's audience and framing resemble the real buying decision
  • the kill threshold — the conversion level, set in advance, below which the candidate is abandoned, so a negative result is a valid, decision-forcing kill
  • the escalation path — the ordering: run the cheap probe first, build only if it clears its threshold, and escalate to progressively costlier probes as cheaper ones pass

What It Is Not

  • Not validation or proof of demand. A smoke test buys signal value relative to test cost, not a verdict — a high click rate is not a validated market. The job is to discriminate real demand from notional demand at a known cost, so reading a pile of sign-ups as "people want this" mistakes a cheap probe for the expensive build it was designed to avoid. The right question is whether this test, at the price paid, could have made the team walk away.
  • Not all signals counted equally. A click, an email, a pre-authorised card, and a deposit are not interchangeable evidence: they sit at increasing cost to the respondent and so at increasing predictive power. The most common smoke-test failure is harvesting a weak interest signal and reporting it as an adoption commitment — only a costly action approximates real buying, so a signal must be discounted to the predictive power its cost-to-the-respondent warrants.
  • Not a prototype or MVP. A smoke test is a false-front with no working product behind it (fake-door, landing page, Wizard-of-Oz) that probes whether to build at all; a prototype or MVP is a partial working build, already past that decision, that measures how the thing performs in users' hands. Treating the two as interchangeable evidence confuses a pre-build demand probe with a post-build performance test.
  • Not a test that can only pass. Falsifiability is the requirement: a probe behind a costless interest button or a leading question is uninformative because a real "no" is not a possible, cheap outcome. The negative result is a result — a landing page with no deposits is a successful kill that bought information, not a failed experiment — and a probe that cannot fail must be redesigned until it can.
  • Not the software-engineering smoke test. The older sense — a quick check that a new build runs at all before deeper testing is committed — shares the cheap-fast-filter-before-expensive-work structure but tests function, not demand. The demand-specific machinery (commitment gradient, conversion-to-demand model, kill threshold on adoption) does not apply to a build-sanity check; the two are distinct lineages that happen to share a name.

Scope of Application

The lean-startup smoke test lives within product and demand-validation practice; its reach is within that practice family, across every tactic that shares its cheap-signal-before-expensive-build structure with the probe artifact swapped. The cross-domain consultation, test-screening, and microdosing cases are genuine co-instances of the parent design composition (signaling + experiment + screening + option_value + falsifiability), which is what carries the cross-domain lesson — not "smoke test" by name.

  • Landing-page and fake-door tests — a described-but-unbuilt product behind a sign-up or "buy" button, the canonical demand probe.
  • Wizard-of-Oz manual back ends — a façade that simulates the product by fulfilling requests manually, testing whether the demand and the workflow are real before automating.
  • Pre-order and letter-of-intent campaigns — collecting deposits or signed intent (Kickstarter pre-commits, B2B LOIs) as a high-commitment demand signal ahead of build.
  • Software-engineering build-sanity smoke test (literal sibling) — the older sense within the same business/technology home: a quick check that a new build runs at all before deeper testing is committed; identical cheap-fast-filter structure, but it filters function rather than demand, so the commitment-gradient and conversion-to-demand machinery do not apply.

Clarity

Calling a probe a smoke test makes legible that its job is signal value relative to test cost, not validation. A team that loses sight of this builds something elaborate to "make sure" before deciding, or treats a pile of clicks as proof of demand; the label reframes the question from "did people like it?" to "did this test, at the cost I paid, discriminate real demand from notional demand?" — and so puts the test's cheapness and its predictiveness on a single axis where they can be traded against each other deliberately.

It sharpens two distinctions the bare word "test" leaves blurred. First, the commitment gradient: a click, an email, a pre-authorised card, and a deposit are not interchangeable evidence — they sit at increasing costs to the respondent and therefore increasing predictive power, and the most common smoke-test failure is to harvest a weak interest signal and read it as an adoption commitment. Naming the gradient lets a practitioner ask exactly what the respondent was made to risk, since only a costly action approximates real buying. Second, it makes legible that the negative result is a result: a landing page with no sign-ups or a pre-order form with no deposits is not a failed experiment but a successful kill that bought information about whether to build at a fraction of the build's cost. The sharper question the concept licenses is therefore not "will this succeed?" but "what is the cheapest probe that could still make me walk away?" — relocating the design effort from building the product to engineering a falsifiable, commitment-bearing signal that precedes it.

Manages Complexity

"Will this product succeed?" is an intractably high-dimensional question — it folds together market size, willingness to pay, competition, channel, timing, and a dozen unknowns — and the honest way to answer it is to build the thing and watch, which is exactly the cost the team is trying to avoid. The smoke test compresses that question to a single tractable scalar: the rate at which a cheap probe converts a commitment-bearing action against a pre-set kill threshold. By engineering the probe to demand a costly-enough response — a deposit rather than a click — the team makes one measured conversion number stand in for the whole demand question, and the build-or-walk-away decision reads off that number rather than off a forecast of the market. The many ways a product could fail collapse to one falsifiable signal placed deliberately on the cost-versus-predictiveness gradient: cheap enough to run now, costly enough to the respondent to be predictive. An analyst tracks a probe's conversion against its threshold and reads the go/no-go directly, turning an unbounded prediction problem into the reading of a single commitment-bearing rate.

Abstract Reasoning

The smoke test licenses inference moves that all turn on the commitment gradient and the kill threshold the compression isolates — reading a respondent's costly action as a predictor of real demand.

Diagnostic — infer real demand from what the respondent was made to risk, not from what they said. The characteristic move is to weight a demand signal by its position on the commitment gradient: a click and a survey "yes" cost the respondent nothing and predict little; an email costs a small disclosure; a pre-authorised card or a deposit costs real money and predicts adoption. From a pile of landing-page sign-ups the move is not to infer demand but to ask what did each respondent actually risk? — and to read a high click rate with zero deposits as notional interest, not validated demand. The most common diagnostic error the concept names is reading a weak-commitment signal as if it were a strong one (harvesting interest and reporting it as adoption); the corrective inference is to discount any signal to the predictive power its cost-to-the-respondent warrants. The move also reads a negative result diagnostically: a probe that drew no commitment-bearing action is not a broken experiment but positive evidence against demand — the absence of a costly yes, at a cost the test was designed to extract, licenses the inference that the demand was notional and the build would have failed.

Interventionist — to learn whether to build, engineer the cheapest probe that could still make you walk away. Because the object is signal value relative to test cost, the licensed intervention is to design the probe, deliberately placing it on the cost-versus-predictiveness gradient: cheap enough to run now, but demanding a response costly enough to discriminate real from notional demand. The move escalates the demanded commitment until the signal is predictive — replace a survey with a landing page, a click with an email, an email with a pre-order or deposit — and sets a kill threshold in advance so the result forces a decision rather than inviting rationalisation. Each design choice is a prediction: raise the commitment demanded and the conversion rate will fall but its predictive power will rise; the analyst's move is to find the point where the test is still cheap yet the surviving signal is strong enough to bet the build on. The master interventionist move is to engineer falsifiability — a probe that can only ever pass (a costless interest button, a leading question) is uninformative and must be redesigned so that a real "no" is a possible, and cheap, outcome.

Boundary-drawing — when does a probe count as a smoke test, and when is the signal trustworthy? The concept draws lines around its own applicability. It separates the pre-build regime — where no product exists and the question is whether to build at all — from the post-build regime, where a real or minimal product is in users' hands and A/B tests or MVP usage measure a different thing (how it performs, not whether to make it). A false-front with no product behind it (fake-door, Wizard-of-Oz) is a smoke test; a partial working build is a prototype or MVP, and reading the two as interchangeable evidence is the boundary error. It also bounds when the conversion number is trustworthy: the signal is only as good as the implicit model mapping probe response to real demand, so the move is to ask whether the probe's audience, framing, and demanded action actually resemble the real buying decision — a deposit from a mis-targeted audience predicts nothing. The concept applies precisely where a cheap commitment-bearing action can stand in for an expensive one; where no costly action is available to extract, the smoke test does not apply and the demand question must be answered another way.

Predictive / order-of-events — the cheap signal precedes and forecasts the expensive outcome. The concept asserts a temporal ordering — cheap probe first, expensive build only if the probe clears its threshold — and predicts that the commitment-bearing conversion rate forecasts the real adoption rate well enough to gate the decision, the prediction improving as the demanded commitment rises toward real payment. The move runs forward (a probe clearing its threshold licenses the build, a probe failing it licenses the kill before any build cost is sunk) and backward (a product that failed in market, reconstructed, often reveals that no costly demand signal was ever extracted before building — the order-of-events fingerprint of having skipped the test). The payoff structure the concept predicts is asymmetric: the probe's cost is small and paid early; the information it buys averts a large build cost paid late, so the expected value of inserting the test before the commitment is high precisely when the demand is most uncertain.

Knowledge Transfer

Within lean-startup and product practice the smoke test transfers as mechanism across every tactic that shares its cheap-signal-before-expensive-build structure, because the design logic is identical and only the false-front changes. The commitment-gradient discipline (a click < an email < a pre-authorised card < a deposit, each costing the respondent more and so predicting more), the kill-threshold-in-advance rule, the falsifiability requirement (a probe that can only pass is uninformative), and the "engineer the cheapest probe that could still make you walk away" move carry intact across landing-page and fake-door tests, Wizard-of-Oz manual back ends, pre-order and letter-of-intent campaigns, and Kickstarter pre-commits — these are one practice with the probe artifact swapped, not analogies. The software-engineering sense of "smoke test" (a quick check that a new build runs at all before deeper testing is committed) is a near-sibling within the broader business/technology home: it shares the cheap-fast-filter-before-expensive-work structure exactly, differing only in what is filtered — function rather than demand — so the cost-versus-information design logic transfers, though the demand-specific machinery (commitment gradient, conversion-to-demand model) does not apply to a build-sanity check.

Beyond that practice family the situation is best read as the third case: the named lean-startup tactic does not travel, but the design composition it instantiates genuinely recurs across distinct substrates as co-instances (case B). Cross-domain probes — policy consultation (green-paper publication smoke-testing political feasibility before a bill commits), course development (MOOC pre-registration before building the curriculum), fundraising (letters of intent before a capital campaign), film and television (rough-cut test screenings before final edit; pilot episodes before a season order), theatre and publishing (out-of-town tryouts before Broadway; preprints and proposal-based contracts), and drug development (Phase 0 microdosing as a cheap pre-Phase-I probe) — are not loose metaphors for a smoke test; each genuinely instantiates the same triple of an expensive commitment, a cheap commitment-bearing probe of whether the commitment is justified, and an escalation rule. But what they share is the abstract design pattern, not the lean-startup tactic with its name and its landing-page/fake-door furniture. The honest move is to carry the pattern, not the named concept.

The reason the pattern travels while the concept stays home is that the smoke test is a composition of substrate-general primes, and it is those parents that recur. Strip the lean-startup codification and the residue is signaling (a costly action revealing hidden information — the theory the commitment gradient instantiates), experiment and falsifiability (a hypothesis test that must be capable of failing), screening (inducing self-revelation through structured choice — the costly-action cousin), option_value / real-options (paying a small early cost to preserve flexibility over a large later commitment), and pilot (small-scale-before-full-commitment). Those parents are genuinely substrate-independent and recur wherever a cheap commitment-bearing action can stand in for an expensive one. The home-bound cargo is everything that makes this the smoke test: the fake-door, landing-page, and Wizard-of-Oz tactics, the build-or-not framing, the conversion-to-demand model, and the lean-startup name itself. The honest report is therefore: across lean-startup practice and its software-engineering sibling the concept transfers as mechanism with only the probe swapped; the cross-domain consultation/screening/microdosing cases are co-instances of the same design composition, which is what should carry the lesson; but "smoke test" as named drags startup-specific vocabulary, so reach for the signaling + experiment + screening + option_value + falsifiability composition instead. (See Structural Core vs. Domain Accent.)

Examples

Canonical

Buffer's fake-door landing page (Joel Gascoigne, 2010) is the textbook smoke test. Before writing any product code, Gascoigne put up a landing page describing the social-media scheduling tool and a "Plans and Pricing" button; clicking it led to a page admitting the product was not built yet, with a field to leave an email address. When that drew enough sign-ups to be encouraging, he inserted an intermediate pricing page listing specific monthly tiers, so a click on a paid plan — a costlier commitment than bare interest — was measured before it too hit the "not ready, leave your email" page. The staged design deliberately escalated what the respondent was asked to risk, and a null result (no clicks, no emails) would have been a cheap, decision-forcing kill.

Mapped back: Building Buffer is the expensive commitment; the two-page façade is the cheap probe; the interest click, the paid-tier click, and the email form a rising commitment-bearing signal along the gradient; staging pricing to sit between free interest and a real email is the cost-vs-predictiveness placement; a possible zero-response outcome satisfies the falsifiability requirement; and the sequence run-interest-then-pricing-then-build is the escalation path.

Applied / In Practice

Phase 0 microdosing in drug development is the same design outside industry-startup practice. Under the FDA's 2006 exploratory-IND guidance (and the parallel ICH framework), a sponsor may give a "microdose" — no more than 100 micrograms, and no more than one-hundredth of the dose expected to produce a pharmacological effect — to a small number of human volunteers to measure how the candidate is absorbed, distributed, and cleared, before committing to a full Phase I trial. The microdose is far too small to treat or (usually) to harm; its job is to extract an early, cheap, human pharmacokinetic signal that discriminates a viable candidate from a loser, so failures are killed before the expensive clinical programme is sunk.

Mapped back: The full clinical-development programme is the expensive commitment; the microdose study is the cheap probe; human pharmacokinetic behaviour is the commitment-bearing signal standing in for eventual efficacy; the ≤100 µg, ≤1/100 dose is the cost-vs-predictiveness placement (cheap and safe enough to run early, yet real human data); a pre-set go/no-go on the PK result is the kill threshold; and microdose-then-Phase-I is the escalation path.

Structural Tensions

T1: Cheap-to-run versus costly-to-the-respondent (the two ends of the gradient pull apart). The probe must be cheap enough for the team to stand up now yet demand a response costly enough to the respondent to discriminate real from notional demand — and these move in opposite directions. Escalate the commitment demanded (a deposit over a click) and predictive power rises, but conversion falls and the test itself grows slower and dearer to build and staff; drop the commitment and the test ships instantly but the signal decays into notional interest that predicts nothing. There is no free placement on the gradient: every point trades the team's run-cost against the respondent's cost-to-act, and the entire design task is choosing that point deliberately rather than defaulting to whatever artifact yields a number fastest. Diagnostic: At the chosen placement, is the demanded action costly enough to the respondent to be predictive, or merely cheap enough for the team to ship?

T2: A probe that can fail versus a probe engineered to pass (falsifiability against the wish for a yes). All the informational value depends on a real "no" being a possible, cheap outcome — yet the team running the test usually wants the product to be worth building, and that wish quietly shapes the probe toward one that can only pass: a costless interest button, a leading question, a hand-picked friendly audience. Such a probe returns a reassuring conversion number that carries no information, because a kill was never on the table. The pull is structural, not a mere lapse of discipline: the same team that decides whether to build also designs the test whose failure would stop them, so the incentive to engineer falsifiability away is baked into who holds the pen. Diagnostic: Could this probe, as built, have produced a cheap, decisive "no" — or was a walk-away outcome never really reachable?

T3: The conversion number versus the model behind it (a signal is only as good as its resemblance to the real buy). The smoke test collapses the demand question to one scalar, but that scalar forecasts real demand only through an implicit conversion model, and that model holds only if the probe's audience, framing, and demanded action actually resemble the eventual buying decision. A deposit harvested from a mis-targeted or artificially primed audience is a costly signal that predicts nothing; a clean number can be perfectly precise and wholly unrepresentative. The bind is that engineering realism into the probe — the right audience, honest framing, the real purchase moment — fights the cheapness and speed that make it a smoke test at all: the more faithful the probe grows to the real decision, the closer it drifts toward the expensive build it was meant to avoid. Diagnostic: Does the probe's audience and framing resemble the real buying decision, or is the conversion precise but drawn from a context that will not recur?

T4: The kill threshold set in advance versus the reading made after the fact (pre-commitment against rationalization). The concept insists the threshold be fixed before the result is seen, so a negative outcome forces a decision instead of inviting reinterpretation. But the team that sets the threshold is the team invested in building, and once a disappointing number lands the pull is to renegotiate — "the audience was wrong," "the framing undersold it," "one more probe." The threshold's whole power is that it binds the go/no-go before motivated reasoning can act, yet nothing mechanical enforces it; the same discretion that set the line can move it. The tension is that a pre-set kill line is only as good as the team's willingness to honour a number they now wish were higher. Diagnostic: Was the threshold written down before the result, and is the team honouring it now — or is the line being relitigated to license the build?

T5: Demand validated versus product unbuilt (a passed probe gates the build but cannot vouch for it). A smoke test extracts a commitment to a described product from a false-front with nothing behind it — which is exactly what makes it cheap and pre-build — but that same emptiness means it can only test the pitch, never the execution. A probe can clear its threshold on genuine demand and the thing subsequently built can still fail to satisfy that demand on quality, workflow, or price-in-practice. Reading a pass as a green light therefore carries a residual the concept is explicit about: validated demand for a promise is not validated performance of a product, and the boundary against the prototype or MVP marks precisely the question a smoke test structurally cannot answer. Diagnostic: Does the passed probe license only "someone wants the described thing," or is it being over-read as "the thing we will build will satisfy them"?

T6: Highest value under uncertainty versus wasted motion under certainty (when the test earns its place). The payoff is asymmetric — a small early cost averts a large late one — but only when the demand is genuinely in doubt; the expected value of inserting the probe is high precisely where the outcome is most uncertain and collapses toward zero where demand is already evident. Run a smoke test on a demand that is obvious and it buys almost no information while signalling indecision and burning calendar time; skip it on a demand that is uncertain and the full build cost rides on a guess. The tension is that the discipline is not costless to invoke, so the same rigor that protects an uncertain bet is overhead on a settled one. Diagnostic: Is the demand this probe would test genuinely uncertain, or is the test buying reassurance on a question already answered?

T7: Autonomy versus reduction (its own lean-startup tactic or the composition of substrate-general parents). "Smoke test" is a named, codified practice with its own furniture — the fake-door, the landing page, the Wizard-of-Oz mock, the conversion-to-demand model, the build-or-walk framing. Yet its portable structure is not proprietary: strip the codification and the residue is a composition of signaling (a costly action revealing hidden information), experiment and falsifiability (a hypothesis test that must be able to fail), screening (inducing self-revelation through structured choice), option_value (a small early cost preserving flexibility over a large later commitment), and pilot (small-scale before full commitment). Those parents are what genuinely recur across policy green-papers, test screenings, and Phase 0 microdosing — co-instances of the same design pattern, not the lean-startup tactic transplanted. Diagnostic: Resolve toward the parent composition (signaling + experiment + screening + option_value + falsifiability) when carrying the lesson cross-domain; toward the named smoke test when designing a demand probe inside product practice.

Structural–Framed Character

The smoke test sits at the framed-leaning position on the structural–framed spectrum — a named, codified lean-startup tactic wrapped around a genuinely portable design composition. The criteria pull framed. Evaluative_weight is modest: the concept is a practice-prescription ("engineer the cheapest probe that could still make you walk away," fix a kill threshold in advance) rather than a neutral mechanism, so it carries normative freight about how a build decision ought to be made, though it renders no verdict on the product itself. Human_practice_bound is high: the tactic presupposes a team facing an expensive build decision, respondents who can be made to risk something, and a market — it is constituted by product-development practice and dissolves without an agent deciding whether to build. Institutional_origin is pronounced: "smoke test" is lean-startup furniture, with its named artifacts (fake-door, landing page, Wizard-of-Oz mock, conversion-to-demand model) and the build-or-walk framing all codified by a business-methodology tradition, not found in nature. Vocab_travels fails outside product practice: the demand-probe machinery (commitment gradient, conversion-to-demand model, kill threshold on adoption) loses its referents off the substrate, and even the software-engineering sibling shares only the cheap-filter structure, not the demand vocabulary. Import_vs_recognize is the subtle case: the design pattern genuinely recurs as co-instances across policy green-papers, test screenings, and Phase 0 microdosing — but that recurrence is recognition of the parent composition, not of "smoke test"; carrying the lean-startup label into drug development imports startup-specific vocabulary the target does not need.

The portable structural skeleton is a cheap, commitment-bearing probe of whether an expensive commitment is justified, run before the commitment and escalated only as cheaper probes pass — cheap-signal-before-expensive-build. That skeleton is substrate-general and is exactly what the smoke test instantiates from a composition of umbrella primes — signaling (a costly action revealing hidden information, the theory the commitment gradient rests on), experiment and falsifiability (a hypothesis test that must be able to fail), screening (inducing self-revelation through structured choice), option_value (a small early cost preserving flexibility over a large later commitment), and pilot (small-scale before full commitment). The cross-domain reach belongs to that composition: policy consultation, test screenings, and microdosing are genuine co-instances of it, while the fake-door/landing-page/Wizard-of-Oz tactics, the conversion-to-demand model, and the lean-startup name that make "smoke test" the specific named tactic stay home. Its character: a practice-prescriptive, practice-bound, lean-startup-codified tactic whose only substrate-spanning content is the cheap-probe-before-expensive-commitment skeleton it composes from signaling, experiment, screening, option_value, and falsifiability, dressed in demand-validation furniture that does not travel.

Structural Core vs. Domain Accent

This section decides why the smoke test is a domain-specific abstraction and not a prime, and it carries the case for its domain-specificity — there is no separate section for it.

What is skeletal (could lift toward a cross-domain prime). Strip the lean-startup codification and a thin relational structure survives: a cheap, commitment-bearing probe of whether an expensive commitment is justified, run before the commitment and escalated only as cheaper probes pass. The pieces that travel are abstract — an expensive, hard-to-reverse commitment held in reserve; a low-cost probe that extracts a signal costly enough to the responder to be predictive; a threshold fixed in advance that turns a negative result into a decision-forcing kill; and a temporal ordering in which the cheap signal precedes and forecasts the expensive outcome. That skeleton is genuinely substrate-portable, which is exactly why the entry can name a whole composition of primes it instantiates — signaling, experiment, falsifiability, screening, option_value, and pilot — each carrying a slice of the structure. But it is the core the smoke test shares with policy green-papers and Phase 0 microdosing, not what makes it the smoke test.

What is domain-bound. Almost everything that makes the concept the smoke test in particular is product- and demand-validation furniture, and none of it survives extraction intact: the false-front artifacts (the fake-door, the landing page, the Wizard-of-Oz mock); the commitment gradient calibrated in the specific currency of demand — a click, an email, a pre-authorised card, a deposit; the conversion-to-demand model that maps probe response to predicted real buying; the kill threshold pinned to an adoption rate; and the build-or-walk-away framing that presupposes a team, a market, and a product decision. These are the worked vocabulary, the instruments, and the empirical cases the practice actually runs. The decisive test: remove the build-or-not decision and the market it answers to, and the machinery has nothing to grip — a landing page collecting deposits is no longer a smoke test but a pre-sale, and the commitment gradient loses its referents. Even the software-engineering homonym keeps only the cheap-filter shape and drops the demand vocabulary entirely, which is why the two are distinct lineages that merely share a name.

Why this does not clear the prime bar. A prime is a relational structure whose vocabulary travels and whose cross-domain transfer is recognition of the same mechanism, not analogy. The smoke test's transfer is bimodal. Within product practice the mechanism travels intact — the commitment-gradient discipline, the kill-threshold rule, and the falsifiability requirement carry across landing pages, fake doors, Wizard-of-Oz back ends, and pre-order campaigns with only the probe artifact swapped, because each supplies the one thing the method needs: a team facing an expensive build and respondents who can be made to risk something. Beyond it the named tactic does not travel; only the design composition it instantiates recurs, as genuine co-instances — policy consultation, MOOC pre-registration, test screenings, Phase 0 microdosing — that share the abstract pattern, not the lean-startup name and its furniture. And when the bare structural lesson is needed cross-domain, it is already carried, in more general form, by the parents: a costly-action-reveals-hidden-information probe is signaling; a test that must be able to fail is experiment under falsifiability; inducing self-revelation through structured choice is screening; paying a small early cost to preserve a large later option is option_value; small-scale-before-full-commitment is pilot. The cross-domain reach belongs to that composition; "smoke test," as named, drags demand-validation cargo that should stay home.

Relationships to Other Abstractions

Local relationship map for Smoke TestParents appear above the current abstraction, mutual partners to the right, and children below. Node labels state whether each abstraction is prime or domain-specific; colors identify relation types.Smoke TestDOMAINPrime abstraction: Optionality — is a decomposition of, conditionalOptionalityPRIMEPrime abstraction: Screening — is a decomposition ofScreeningPRIMEDomain-specific abstraction: Problem-Solution Fit — is part of, conditionalProblem-SolutionFitDOMAIN

Current abstraction Smoke Test Domain-specific

Parents (2) — more general patterns this builds on

  • Smoke Test is a decomposition of, conditional Optionality Prime

    The probe cost can act as an option premium that caps early downside while preserving the right, without the duty, to make the larger build commitment later.

  • Smoke Test is a decomposition of Screening Prime

    The uninformed product team designs a structured choice that induces people who privately know their willingness to commit to reveal it through their action.

Children (1) — more specific cases that build on this

  • Problem-Solution Fit Domain-specific is part of, conditional Smoke Test

    A fake door, pre-order, or landing-page Smoke Test can be one cheap need-side instrument inside a Problem–Solution Fit gate before the product is built.

Hierarchy paths (4) — routes to 4 parentless roots

Not to Be Confused With

  • Minimum viable product (MVP) / prototype. A partial working build placed in users' hands to measure how the thing performs — already past the decision to build. A smoke test is a false-front with no product behind it that probes whether to build at all. Tell: is there a real (if minimal) working product being used (MVP/prototype) or only a landing page, fake-door, or Wizard-of-Oz façade with nothing behind it (smoke test)? Pre-build demand probe versus post-build performance test.
  • A/B test. A controlled comparison of two variants of a live product to see which performs better on some metric — a post-build optimization tool. The smoke test runs before any product exists and asks a binary build-or-walk question. Tell: is the test comparing variants of something already shipped (A/B) or extracting a commitment signal for something not yet built (smoke test)? A/B optimizes an existing thing; the smoke test decides whether the thing should exist.
  • Software-engineering smoke test (the homonym). The older sense — a quick check that a new build runs at all before deeper testing is committed. It shares the cheap-fast-filter-before-expensive-work structure exactly but filters function, not demand, so the commitment gradient, conversion-to-demand model, and adoption kill threshold do not apply. Tell: does the check verify that code executes (build-sanity smoke test) or that customers will buy (lean-startup smoke test)? Same name, distinct lineages — function versus demand.
  • Concept test / market survey. Asking prospective customers whether they would buy — a cheap but low-commitment, easily-passed probe. It sits at the weak end of the commitment gradient and often fails falsifiability (a survey "yes" costs nothing). A smoke test deliberately extracts a costly action (deposit, pre-authorization) precisely because stated intent is not predictive. Tell: does the respondent merely say they want it (concept test) or risk something to get it (smoke test)? The gradient position — what was actually risked — is the difference.
  • Pre-sale / pre-order fulfilment. Collecting real payment for a product you do intend to deliver — a genuine sale, not a probe. A pre-order form used only to gauge demand is a smoke-test tactic, but once deposits are taken toward an intended build it has become a pre-sale. Tell: is the deposit an information-extracting signal on a maybe-build (smoke test) or a committed transaction on a will-build (pre-sale)? The entry notes a landing page collecting deposits toward a planned product is no longer a smoke test but a pre-sale.
  • The parent design composition (signaling + experiment + screening + option_value + falsifiability + pilot). The substrate-general pattern — a cheap commitment-bearing probe of whether an expensive commitment is justified — that the smoke test instantiates. It is what actually recurs cross-domain (policy green-papers, test screenings, Phase 0 microdosing); "smoke test" is the lean-startup instance. Tell: strip the fake-door/landing-page/conversion-to-demand furniture and what remains — costly-signal-before-expensive-commitment with a kill threshold — is this composition, not "smoke test" by name. (Treated fully in a later section.)

Neighborhood in Abstraction Space

Smoke Test sits in a moderately populated region (53rd percentile for distinctiveness): it has near-neighbors but no dense thicket of look-alikes.

Family — Lean Validation & Startup Signal Theater (8 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-07-12