Glossary

A claim about the world that the plan depends on, but that has not yet been tested. Most startup failures trace back to one untested assumption.

A product that builds an interface, workflow, or service around one or more third-party AI models. The label describes an architecture; it does not decide whether the product is useful or defensible.

A company whose product depends on one or more generative or predictive AI models for its core output. Validation must cover workflow value, output quality, model-layer risk, and defensibility — assumptions that do not appear in conventional SaaS validation.

Testing a developer-facing API through a real integration workflow before building a broader product around it.

The set of people who influence, approve, use, or pay for a business product, rather than a single decision-maker.

A small piece of software that runs inside a web browser and modifies or extends a page's behavior. Validation challenges are distinct from SaaS: the user is one click away from disabling the extension, the Chrome Web Store is the dominant discovery channel, and cross-browser maintenance is a real cost.

A manual version of the product experience delivered directly by the founder so a key assumption can be tested before automation is built. The customer pays for an experience, not for code; the founder does the work by hand.

The ongoing process of finding and talking to the people who have the problem you want to solve, before and after you build. Discovery is where most of the actual learning happens.

The single assumption that, if false, would invalidate the rest of the plan. Finding and testing this assumption is the highest-leverage work in early-stage validation.

The rate at which existing customers stop paying. A SaaS business works when new customers per month exceed churned customers per month, and the gap is growing.

A 25–30 minute structured conversation with a person who has the problem, in which the founder asks open-ended questions about the customer's life and workarounds and pitches nothing.

The tendency to interpret interview answers, metrics, and feedback as support for the founder's existing hypothesis, even when the evidence is mixed or contradictory.

The evidence that a defined group of people has a real, recurring problem and is willing to spend time or money to solve it. Demand is observed in behavior (signups, payments, repeat use), not in opinions, and it is the prerequisite for any willingness-to-pay test to produce a useful signal.

The total revenue a single customer generates over the entire relationship with the product. Usually reported as cohort LTV (cumulative revenue per cohort over 12 or 24 months). The LTV:CAC ratio is the financial translation of product-market fit.

The fully loaded cost of acquiring one paying customer, including ad spend, sales time, tooling, and content production. Used as the denominator in the LTV:CAC ratio that measures PMF.

The percentage of a defined cohort (users who joined in a given week or month) who remain active at later time periods. Plotted as a curve. Cliff, slope, or plateau. The plateau shape is the visual signature of product-market fit.

A real, observable instance of a customer losing money, time, or opportunity because of a problem. Customer pain is distinguished from a customer need by the presence of action: pain produces workarounds; need often does not.

A repeatable path that moves a stranger from first contact to paying customer. Examples: SEO, Reddit, Product Hunt, cold email, paid ads, communities.

The conditions that make a product harder to copy or replace after competitors gain access to similar models — such as distribution, workflow integration, permissioned data, trust, or switching costs.

In a two-sided marketplace, the participant group that buys, hires, or consumes what the supply side provides. Marketplace founders must validate the demand side separately because demand intent depends on the supply side being useful and reachable, not just on the buyer's interest.

An early customer who works with a startup to shape a product around a real workflow before broader sales or distribution.

The probability that the founder cannot reach the target audience at a cost the unit economics tolerate. Distribution risk is named separately from market risk because the audience may exist and want the product, and the founder may still fail to reach them at a viable CAC.

The discipline of specifying the hypothesis, treatment, measurement, and decision of an MVP test before the test runs. A founder who designs the MVP as an experiment — instead of as a launch event — produces evidence rather than opinions.

The path by which a browser extension reaches users, including the workflow, permission boundary, and channel where it is discovered.

The first person who pays for the product, not the first person who says they would. The first customer is the unit of evidence that matters in pre-revenue work.

The overlap between your background, skills, audience, and domain experience, and what this specific idea needs. Some ideas need a domain expert. Some need an audience. Few need neither.

A bounded period during which a customer can use the product without paying. A free trial is a real validation tool, but it tests curiosity, not willingness to pay, and complicates churn math.

A customer signal that looks like evidence of demand but is actually politeness, hypothetical enthusiasm, or curiosity — anything that could not have produced a believable no.

The narrow description of the buyer most likely to buy, retain, and refer. Not a segment — a specific kind of customer with a specific budget and a specific workflow.

A small experiment in which you publish a single page describing the product, drive a small amount of traffic, and measure how many visitors take the next step (sign up, request access, leave an email). Cheap signal, not proof.

The smallest version of the product that lets a team run a complete Build–Measure–Learn loop with the least amount of effort. The MVP is a learning tool, not a finished product; its output is evidence about an assumption, not revenue.

A small, often solo-built SaaS that targets a narrow use case, serves a small audience, and is typically priced below $50/month. Validation challenges are the same as larger SaaS; the cost of being wrong is smaller in dollars but larger in proportion.

The evidence that a defined group of people has a real, recurring problem and is willing to spend time or money to solve it. Distinct from product validation, which tests whether the specific product fixes the problem.

A durable advantage that protects customer relationships or margins and can strengthen over time. For an early startup, a moat is a hypothesis to test, not a claim to announce.

Exposure to upstream changes in model behavior, price, limits, availability, or terms that can alter the product without the founder changing its own code.

The practical likelihood that a participant can find a suitable counterpart and complete a useful transaction in the market.

The share of installs that open the app again on day 1, day 7, and day 30 after install. Day-1 retention is the cheapest signal that the app has a habit trigger; day-30 retention is the signal that decides whether the app survives the standard mobile retention cliff.

New customers or signups that arrive without paid acquisition — through SEO, direct navigation, word-of-mouth referral, or community sharing. A meaningful organic share of total growth is one of the four signals of product-market fit. Organic growth below 20% of total is a signal that fit has not yet been earned.

A clickable or interactive model of a product used to test the workflow and the design before building. A prototype is for the team; an MVP is for the customer. The prototype answers 'does this work as designed?' — the MVP answers 'does this work in the customer's life?'

The point at which the product satisfies a real, repeated demand from a specific audience. It is observed, not declared — you know you have it when retention and word-of-mouth improve without paid acquisition.

A conversation with a potential customer about their life and workarounds, not about your product. The interview is your chance to learn what they actually do, not what they say they would do.

The evidence that real buyers will pay the planned price — not a discounted launch price, and not a hypothetical interview answer. Usually tested with presales, deposits, or a 30-day concierge at the business-model price.

A small group of customers who pay the founder, in advance, for an early version of the product, with the understanding that they will give feedback and the founder reserves the right to refund. A paid pilot is the strongest pre-build pricing signal because the customer has done the thing that actually matters: given money.

The state in which a defined customer has a real, recurring problem that a defined solution would meaningfully improve — before either the customer or the founder has committed to building or buying anything. Observed in customer behavior (workarounds, paid alternatives, time spent), not in interview agreement.

A specific, observable instance of customer pain — the moment the customer loses money, time, or opportunity because of the problem. Distinct from the general problem: a problem is the category, a pain point is the specific instance that produces action.

The share of users or customers who remain active over time. Measured as a cohort retention curve at day 1, day 7, day 30, and beyond. The single most informative PMF signal: a curve that flattens indicates fit; a curve that declines steadily indicates the absence of fit.

The discipline of testing the assumptions behind your idea — the problem, the customer, the willingness to pay, and your ability to execute — before you commit months of work to building it.

A small experiment that measures whether people take a meaningful next step toward a proposed product before the product exists.

Software sold as a recurring subscription rather than a one-time purchase. The customer relationship is continuous, so the product has to justify itself every month.

The SaaS-specific subset of startup validation that tests the recurring-revenue assumption stack: recurring willingness to pay, churn past first renewal, ICP narrowness, pricing-tier fit, and distribution-channel reachability. Generic validation advice tests first-payment intent and skips the renewal question; SaaS validation does the opposite.

In a two-sided marketplace, the participant group that provides the listed good, service, or labor. Marketplace founders must validate the supply side separately from the demand side because each side has its own recruitment, retention, and unit-economics reality.

A second-round conversation after the founder has a proposed solution, testing whether the proposed solution actually solves the problem that the problem interview surfaced.

The percentage of respondents to a one-question survey who say they would be 'very disappointed' if they could no longer use the product. The threshold proposed by Sean Ellis in 2010 is 40%. Treat as a directional signal of stated attachment, not a verdict on behavior.

A specific, testable claim about what the proposed product would do for a defined customer. Testable with interviews and concept tests; the customer who can describe what would change if the problem disappeared has named a solution hypothesis, not a feature list.

A market that serves two participant groups whose ability to transact depends on the other side being useful and reachable.

Rob Fitzpatrick's 2013 rule that you should never ask people if they would buy your product, because they will lie to be polite, and instead ask them about the last time they tried to solve the problem your product would solve.

How often and how deeply users interact with a product over a measured period. Engagement precedes retention: a user who never logs in cannot be retained. The right measurement is frequency × depth (sessions per week × core actions per session), not raw session count.

What the user is missing in their current workflow — the gap between what they have and what they would need to solve their problem. A need that does not produce action is not a pain; pain severity, not need enumeration, is what problem-solution fit tests.

Run cheap experiments that test the problem, customer, and willingness to pay before committing to a build. Validation produces evidence, not certainty.

A specific, falsifiable test designed to produce evidence about one assumption in the validation framework. The smallest experiment that can produce a believable signal is the right one; the cheapest method that can test the assumption is the right method.

The specific change the customer would experience after using the product, expressed in the customer's language and tied to the alternative they currently use. A value proposition is testable, not a tagline — it should make a customer able to say yes or no without further explanation.

An MVP that looks automated to the customer but is operated manually behind the curtain. The customer experiences the product; the founder (or a hidden human) produces the output. The signal is whether the customer would pay for the experience if it were automated at scale.

Whether a real person will trade real money for the solution, in a real transaction. Only money tests money — interviews and surveys do not.

New customers acquired because existing customers told other people about the product, without being incentivized. The second of four PMF signals. Word-of-mouth without a referral program is the cleanest behavior signal that the product solves a recurring problem.

What the customer does today to address the problem — a spreadsheet, a freelancer, a tool, a process, an abandoned workflow. Workarounds are the cheapest signal of real pain because they are observable behavior, not opinion. The workaround's cost is the price anchor for any future solution.