The vocabulary of product, defined once.

Concrete examples of the finished behavior that a stranger could check, written as things a person does, not qualities the product has. Also the test plan for the goal the work serves.

Getting users, measured by cost per activated and retained user, not cost per sign-up. Find the one channel that works before the many that might; weight the ones that could close a loop.

The share of new users who reach the first real value, the moment the product proves itself. Not sign-up, which is a promise. Read as a cohort, never a snapshot.

Four values and twelve principles about shipping working software and responding to change. In this book, read as the page, not the industry that grew around it.

The specific early action that separates users who stay from those who leave. Found by comparing cohorts, then made the definition of activation and the target of onboarding.

A product owner whose material is model behavior: prompts, evals, context and the boundary between what the model does and what the product decides.

The time a team is willing to spend on a piece of work, fixed before the work begins so the work is shaped to the time rather than the time to the work.

The first customer: the overlap of the segment that wants it most, the segment that will pay, and the segment the first version can serve. Small on purpose; size is the wrong criterion.

A strategy statement of the form we will and we will not, in dependency order, with a critical question: what has to be true for this to hold, and one number to check it in ninety days.

The step that limits how fast a team learns what to build. It moved from building to gathering what real people do and having the discipline to let it overrule the team's vote.

A requirements document written to be uploaded, not read: named personas, the rules already decided, a paste-ready dataset, the screens allowed, what is out. Its reader forgets nothing and invents anything you leave blank.

A rough prototype is a way of understanding, not a way of delivering. Make the idea tangible early and let the reaction teach you what you meant.

The loop at the heart of Lean Startup. Its output is not a product but learning, and the learning is the input to the next turn. A loop, not a phase.

Nine boxes; for a first version, the ones that decide viability are revenue streams, cost structure, and what the team owns versus borrows. A model built on a partner's asset is a model the partner can end.

A group of users who started in the same period, followed over time. The smallest honest unit of after-launch learning; a snapshot mixes newcomers with veterans and tells you nothing.

In a conversation, something the person gives up that costs them: a second meeting, an introduction, time, a document. In strategy, a bet the team writes down and is bound by.

Where the problem is discussed without you in the room. You are collecting two things: the words people use, and the workarounds they are proud of. Listen before you speak; never pitch.

Delivering the value by hand, in the open, for a few customers, before building anything. The richest rung of the fidelity ladder, because you are inside the customer's operation.

What a product owner becomes when every document is generated from one operating brief. The job shifts from producing documents to maintaining the context that produces them.

Teresa Torres's habit of a weekly customer touchpoint. Discovery is not a phase that ends but a capability the team keeps, before and after launch.

What has to be true for a bet to hold. Written next to the bet; a bet whose question you cannot write is a wish with a sequence number.

A conversation about the person's past, not your idea: what happened last time, what they did, what it cost. Past behavior over opinion, commitment over compliment.

A short record of what the evidence said, what was decided and why, what was traded off, and what the team learned about how it decides. Written so the reasoning survives the decision.

A builder who designs and codes the same artifact, collapsing the old handoff between the two. A role the AI era made common at small companies.

IDEO and the Stanford d.school's human-centered method. Its lasting gifts are empathy as the starting point and building to think; its cost at startup scale is the workshop it wraps them in.

Do people want it. Proven by commitments and behavior, not by enthusiasm.

Reading what the world already knows: the market, the competitors, the adjacent tools, and the post-mortems of companies that tried before. Someone has tried this; find out what happened.

Once a decision is made, those who argued otherwise say so, on the record, and commit. The record is what makes the commitment real and keeps the dissent from returning as sabotage.

Rita McGrath and Ian MacMillan's method: list a venture's assumptions, attach numbers, and test the one that would kill the plan first.

Open to understand, close to decide, open again to test, close to conclude. The spine of the method; the closes matter more than the opens.

Observation before opinion: you do not know the problem until you have watched someone have it. Design Thinking's most durable contribution to product.

What you learn when something is in a stranger's hands: where they got stuck, what they expected, what they agreed to at the end. A yes, a vote or a compliment is not evidence.

Five rows rated honestly before a direction is chosen: problem validation, market timing, solution fit, team capability, resource match. Each with its evidence and the signal that would change it.

The smallest real thing shipped and learned from. The part of the owner's job that turns a vision and a strategy into a first unit of work.

How many complete, honest experiment loops a team finishes per unit of time. A growth metric in its own right, because learning compounds. Honest means a real hypothesis, threshold and decision, not more tests on trivia.

An hour with someone who spent years inside the problem: a former operator, an analyst, a category veteran. Ask for help, not validation, and use the last fifteen minutes for introductions.

An entry point for a feature that does not exist, placed in the live product to measure how many users try it. The cheapest way to test demand, and it must resolve honestly.

Can we build and deliver it with the team, the partners and the operations we actually have. A working demo answers none of this.

The one-page plan a team takes into interviews: what is being validated, the question bank, the opening words, the tasks as situations, the commitment asks, and who speaks and who writes.

Reasoning from what is actually true rather than from analogy or convention. Diagnosis before policy; the base facts before the plan.

The one-month column of the roadmap read as a first version: scope in and out, use cases, acceptance criteria. A consequence of the ladder, not a vote.

Jesse James Garrett's model of experience from the abstract down: strategy, scope, structure, skeleton, surface. Each plane is decided before the one above it can be right.

An engineer embedded with the customer, building against their real problem in the open. The concierge rung of the fidelity ladder, staffed by someone who can ship.

One sentence, one napkin, thirty seconds, current resources. Cheap checks that reveal whether a solution is clear enough that testing it would mean anything.

Acquisition, activation, retention, revenue, referral. A diagnostic that describes a business that leaks: each stage keeps a fraction. Used to find the weakest stage before running growth experiments.

One number per bet, with a threshold and a date: what must we see to keep believing this? Below the number, the bet is in doubt, not the team. Every goal has a parent bet.

A system where the output of one cohort becomes the input of the next: content, invitations, data. A funnel leaks and must be refilled; a loop compounds and feeds itself.

The middle of Rumelt's strategy kernel: the overall approach that turns a diagnosis into coherent action. In this book, the 'we will and we will not' that makes a strategy a choice.

Google's categories for measuring experience: happiness, engagement, adoption, retention, task success. Used to choose the few honest measures a small team can read by cohort.

The order of research: secondary before experts, experts before communities, communities before customers. Each source is cheaper than the next and makes the next better.

A statement that can be wrong, in the form if we do X then Y will happen, measured this way, because Z. If you cannot say what result would prove you wrong, you are looking for applause.

A prioritization score for experiments: impact, confidence, ease. Its value is less the arithmetic than the conversation, which surfaces the easy-but-pointless tests.

Eric Ries's term for the record of what a team has learned. At one-to-five scale, one searchable entry per experiment, so learning compounds instead of evaporating.

Understanding earned from what people do rather than what they say, synthesized until patterns outrank anecdotes and the problem can be stated in the customer's own words.

Information is what you gather; insight is the pattern or contradiction you notice in it. Most teams have plenty of the first and starve for the second.

Measuring what the experiments need and nothing else: the funnel as events, users tagged by cohort and channel, the activation and core-retention events, verified by hand. Over-instrumentation buries the signal.

A test for a unit of work: independent, negotiable, valuable, estimable, small, testable. A story that fails 'small' is two stories; one that fails 'testable' is a wish.

The decision the evidence framework forces. Iterate when the core is validated and the solution needs work; pivot when a fatal row is red. The one thing not allowed is neither.

Clayton Christensen's frame: people hire a product to make a specific progress in a situation. A job is a progress, not a demographic or a feature request.

The customer side of the value proposition canvas: the progress they are trying to make, what makes it hard, and what they hope for. Written from quotes, not imagination.

Mission, vision, strategy, goals, roadmap, task. Each answer is the input to the next; a vague level turns everything below it into opinion.

Smoke test, concierge, Wizard of Oz, navigable prototype. Each answers a different question at a different cost. Build the cheapest thing that can be wrong.

Eric Ries's discipline of validated learning through build-measure-learn loops, with the minimum viable product as the instrument. The lineage behind experimentation.

One place where key quotes, surprising findings, patterns and open questions accumulate across a sprint. Its value is that the whole team, including whoever missed the interview, sees the same evidence.

Places to intervene in a system, ranked by how much they move it. The highest are rarely the loudest, which is why teams push on the wrong ones.

Deciding well before the evidence is in, by naming what you do not know, its consequence and its likelihood, and testing the unknowns that matter in the order they would hurt.

A named way of working with a lineage: Design Thinking, Jobs to Be Done, Lean Startup, Agile, the Double Diamond. The book credits each one, says what it assumes, and where the assumption fails at startup scale.

The three commitments that make an experiment trustworthy, all made before it runs: one metric, one number that means yes, and what the team will do at each outcome.

Numbers are instruments, not steering. At a startup with almost no data, a dashboard that drives strategy is driving it off noise.

The step between a written idea and something a stranger can use. For twenty years the most expensive step in the loop; AI collapsed it, which is why the bottleneck moved to the two human ends.

The smallest thing that lets the team learn what it most needs to learn. Defined by its purpose, not its size; often a concierge or a landing page rather than a small version of the product.

What the company exists to do, in one sentence, true at every point in its life. The one level of the ladder made from conviction rather than evidence.

Rob Fitzpatrick's rules for customer conversations: talk about their life, not your idea; specific past events, not hypotheticals; listen more than you talk; seek a commitment, not a compliment.

Several fronts, each a hypothesis with a number and a date: who pays, for what, how much, how packaged. Pricing is a discovery question, not a launch question.

Sorting work into must-have, should-have, could-have and won't-have. Fast, and its 'won't' column is where the excluding happens.

Screens a stranger can tap through on their own device, with realistic data and nothing behind them. The rung AI changed most: two days from a written brief, by anyone.

Whether a cohort's total revenue grows or shrinks over time after churn and expansion net out. A cohort whose expansion outpaces its churn grows revenue with zero new customers.

The single measure that best captures the value the product delivers, with a few input metrics that drive it. Chosen once the product exists and the loops are running.

Horizons instead of dates on a roadmap, because dates on a startup roadmap are fiction and horizons are honest. What is in development, what is planned, what is direction.

Jeff Bezos's distinction: a two-way door can be walked back cheaply and should be decided fast; a one-way door cannot and deserves care. Sequence the reversible early and the irreversible late.

The rule the whole AI-era method rests on. When a feature costs one prompt, nothing rations features but discipline. A feature without a question behind it has no way to fail correctly.

The one document an engagement is generated from: the cast, calendar, deliverables, voice, decisions and rules. Everything else is derived from it and regenerated when it changes. Upload only the brief.

An outcome at the top, opportunities beneath, solutions beneath those. A way to keep every solution traceable to a problem worth solving.

Generate many solutions before picking one. The first solution is the one everyone thought of; divergence, done in silence and stealing from adjacent industries, finds the others.

For each key resource, whether the company owns it or borrows it from a partner. Value created by a borrowed asset belongs to whoever owns the asset. Build the owned asset first.

Ask what happened the last time, not whether something would be useful. The customer is an expert on their past and an amateur about their future. Interview the expert.

Evidence you gather yourself, from experts, communities and customers. The most expensive source, spent last, once you know what to ask.

A rule the method depends on. Each principle in the book exists because of a specific moment where its absence cost something, and that moment is told next to it.

One table scoring every problem cluster on the same four things: frequency, pain, current solution, willingness to pay. Lets a pattern outrank the anecdote you happen to remember best.

Needs, behaviors, workarounds and constraints as they exist before you build anything. You can photograph a problem; you cannot photograph a solution.

Five lines in the customer's words with a number attached: for whom, in what situation, why it matters, what they do now, and how that fails. The output of the sprint's first half.

The steps in order, and what each one hands to the next. In this book a process is always described as a flow of increasing certainty, never as a workshop agenda.

At an early-stage startup, not a role but a responsibility: the ownership of strategy, vision and execution at once, held by one accountable person on a team of one to five.

An engineer who owns outcomes, not tickets: talks to users, decides what to build, and ships it. One of the roles product blurred into once building got cheap.

The one person answerable for what gets built, why, and in what order. The work is done by everyone; the accountability sits with one.

The shortest path from an assumption to a written position tested against people who owe you nothing. Two halves: insight is earned, then evidence is tested.

At small scale, three behaviors together: a retention curve that flattens, activation that repeats, and a payment from the right segment. The green light for growth.

What has to be true for a roadmap item to move from later to next, or next to now. The goal restated as a gate. Deferred is not discarded when the criterion is written down.

A question made tangible, not a first draft of the product. Write the questions before you build; only build what came from one. The silence while a stranger looks for the button is the finding.

Who is responsible, accountable, consulted and informed for a decision. For a team of five it fits in one paragraph, and its point is that one person is accountable for each thing.

Turning users into a channel: a user's use of the product produces the next user. The purest growth loop, and it only works on a product people already love, so it comes last.

Whether people come back, and the foundation everything else rests on. The container; growth is the water. Nothing else matters until the curve flattens above zero.

A cohort's share still active over time. A curve that decays to zero is a leaky bucket; a curve that flattens above zero is the first real evidence of product-market fit.

Start from the profit the unit must return for the business to make sense, and work backwards to what each line must be, turning forecasts into hypotheses to test.

A prioritization score: reach, impact, confidence, effort. Honest about confidence, which most methods hide. Use one prioritization method, not three.

The assumption that would kill the product if it were false. Experiments are ordered by consequence times uncertainty, not by what is easiest to test.

What you build to move the goals, in horizons with a promotion criterion for each jump. Every item reads 'can we do X to move Y'; if there is no Y, it is not on the roadmap.

One page per interview, filled before leaving: verbatim quotes, where they got stuck, what they expected, what they agreed to and with what face. By morning the words remain and the evidence is gone.

Evidence that already exists: reports, reviews, pricing pages, funding news, failed-startup post-mortems. Done first, because it answers the questions you should not waste a human on.

A group described tightly enough to say who is out. 'Small business owners' is not a segment; 'owners of two to five restaurants who count inventory by hand' is.

Give a test task as a moment in the person's life, not a command. 'You just sat down, show me what you would do,' never 'tap check-in.' An instruction tests your labels; a situation tests your product.

A landing page, ad or email describing a thing you have not built, with a way to say yes. Answers whether anyone wants it. A sign-up is a compliment with an email address attached.

A founder building alone, now more than a third of new startups. The sprint and the whole method are written so one person can run them; the absence of a partner is survivable, the absence of anyone to argue with is not.

A one-page description of one candidate solution that forces its assumptions into the open: concept, journey, requirements, fit with the problem, and risks. Built for comparison.

The options for answering a validated problem: what you could build, how feasible it is, whether it closes as a business. Many options before one.

How fast information moves from wherever it landed to whoever decides. At a five-person startup, the scarce resource, and the owner's real job is to keep it above the speed of building.

Whether a difference is distinguishable from chance. At startup scale most samples cannot support it, so a difference of a few users is a coin flip, and the honest tool is a labeled qualitative read.

A short chain of choices from what the company exists to do down to what it builds first, each level saying what it excludes. A strategy that does not exclude is a wish.

A ladder filled from the evidence and cited, presented as a draft to be corrected. People correct something concrete far faster than they fill a blank.

Deciding, while it is still true, what you just saw. It happens after each conversation, weekly, and once at the end. Patterns over anecdotes; same-day, or the evidence is gone by morning.

Seeing a problem as a set of connected flows and feedback rather than isolated parts. Donella Meadows' primer, and her ranking of where to push.

Deep in one discipline, literate across the others. The shape of a person who can hold product at a startup, where nobody can afford to own only one thing.

The cost of a shortcut taken to ship, recorded in a ledger with the shortcut, the date, why it was taken and what it costs to unwind. Read before planning; pay down what is due, let the rest ride.

Every assumption sorted into validated, open or discarded, with the quote and the interviewee that moved it. The written position a sprint hands to strategy. Open is a state; undecided is not.

Asking a person to narrate what they see, expect and look for while they use the product. The single most valuable usability method, and it costs nothing.

Sorting evidence of value into would pay, would never pay, and gave freely. What people give freely you cannot charge for and must not pay for; it is where a business model quietly breaks.

A short document that proposes, not decides: act one, what ships first because it depends on no one; act two, what follows; what is deferred and why it is not discarded.

What one unit of the core mechanic costs and returns over a defensible period. If one unit loses money, scale multiplies the loss. The AI cost line is the one teams underestimate.

What using the product is actually like, across the five planes from strategy to surface. Not the interface alone: the whole path from why the product exists to the pixel a stranger taps.

The surface a person touches: the screens, controls and words. The last of the five planes, and the only one most teams argue about, which is why the book argues about the other four.

As a [person], I can [action] so that [outcome], with acceptance criteria attached. A unit of work framed from the person's side, not the system's.

Peter Morville's seven facets of a good experience: useful, usable, findable, credible, desirable, accessible, valuable. Each answered by watching one stranger, not by the team.

Progress measured not by features shipped but by assumptions confirmed or refuted against behavior. The unit of progress in the Lean tradition.

The thing a price scales with. Chosen to grow with the value the customer receives, it feels fair; chosen wrong, it grows into resentment and churn. Often matters more than the price level.

One per segment: the customer's jobs, pains and gains on one side, the product's relievers and creators on the other, with the fit written in the customer's words. A proposition for everyone is for no one.

A number that goes up with effort but does not connect to value or revenue: total sign-ups, page views, time in app. It always rises, which is why it feels like success and is not.

Does it close as a business: the model, the unit economics, who pays and how much. The lens startups postpone and the one that comes back.

The description of the world when the company has done its job. Held stubbornly: it changes only when evidence about the world, not a failed feature, contradicts it.

One list of every business-design assumption as a falsifiable claim with a number, evidence, an owner and a date, sorted so the assumption most likely to kill the business is tested first. The deliverable of business design.

Whether people pay, shown rather than stated. Stated is a compliment; revealed is a card being charged. Tested early, with real money, not decided at launch.

A real-looking interface with a person behind it. Answers whether the experience works before the machine that powers it exists. In the AI era, often a general model pretending to be a specialized product.

What a sprint ends in: a tension map and a short memo, short enough to read in one sitting and specific enough to be wrong. The first thing strategy reads.