glossary
The vocabulary, defined plainly.
· A/B testA controlled experiment that splits users into two groups and measures which variant performs better on a pre-specified metric. metrics → · Acceptance criteriaThe specific, testable conditions a user story must satisfy before the PM signs off. general → · ActivationThe percentage of new users who reach a defined first-value moment within a specified time window. metrics → · AgileAn iterative operating model that ships in short cycles, learns from users continuously, and adjusts priorities based on evidence rather than upfront plans. general → · AI agentA system that combines an LLM with memory, planning, and tool access, running in a goal-directed loop until it hits a stopping condition or a guardrail. ai → · AI hallucinationA confident, plausible-sounding model output that is factually wrong or unfaithful to its source material. ai → · APIA contract between two systems specifying how to exchange data, where neither side needs to know how the other is built internally. technical → · Burndown chartA sprint or release chart showing remaining work (Y axis) versus time (X axis), with an ideal line and actual line, used to answer "will we finish?" general → · Canary releaseA deployment strategy that routes a small, deliberate slice of production traffic to a new version to detect regressions before full rollout. engineering → · Churn rateThe percentage of customers (or revenue) lost in a given period, measured against the starting count. metrics → · Cohort analysisGrouping users by a shared starting point, then tracking what share return over time, to diagnose retention rather than just measure it. metrics → · Context windowThe total token budget a model can see and produce in one inference call, shared across input and output. ai → · Conversion rateThe fraction of users in a defined eligible pool who complete a specific intended action in a given time window. metrics → · Customer acquisition cost (CAC)Total sales and marketing spend divided by new customers acquired in the same period. metrics → · DAU/MAU ratioAverage daily active users divided by monthly active users, expressed as a percentage. A stickiness proxy that measures how often users return within a month. metrics → · Design thinkingA non-linear, iterative problem-solving approach that starts with deep user empathy before generating and testing solutions, used by PMs to frame problems before committing to a direction. general → · DogfoodingUsing your own product in real work, not demos, to surface issues before external users encounter them. discovery → · EmbeddingsNumerical representations of meaning that allow retrieval by semantic similarity rather than exact keyword match. ai → · EvalA structured test that checks whether an AI system's output meets a defined quality standard, run before or during production. ai → · Feature flagA runtime configuration switch that controls whether a feature is visible to users independently of whether the code is deployed. engineering → · Fine-tuningContinuing to train a pre-trained model on a curated dataset so its weights update, changing behavior, format, or narrow-task accuracy rather than adding new knowledge. ai → · FunnelA sequential model of the steps users take from entering a product context to completing a defined goal, where volume narrows at each step as users drop off. metrics → · Go-to-market strategyThe plan that determines who you sell to first, how they find and buy the product, and what evidence proves it is working before you scale. strategy → · Ground truthThe verified, authoritative reference data against which a model's outputs are compared to determine whether they are correct. ai → · Human in the loopA design pattern where a human must approve or correct an AI output before the system takes action, as distinct from human-on-the-loop (monitoring with the option to interrupt) and fully autonomous operation. ai → · Jobs to be done (JTBD)A framework holding that customers hire products to make progress in a specific situation, not to own a product or use a feature. discovery → · Latency vs throughputLatency is how long one request takes to complete; throughput is how many requests the system handles per unit of time. Improving one often degrades the other. technical → · LTV (lifetime value)Cumulative gross profit a customer generates from first purchase until they churn. metrics → · MicroservicesAn architecture where an app is split into small, independently deployable services, each owning its data and communicating via APIs. technical → · Model Context Protocol (MCP)An open standard, now governed by the Linux Foundation, that lets AI clients and tool servers connect through a single shared protocol instead of one custom integration per pair. ai → · Model driftThe umbrella term for AI model performance degradation caused by shifts in input data, the real world, or what counts as a correct output. ai → · MonetizationThe set of decisions that determine how a product captures a share of the value it creates: which customers pay, for what unit, at what price, and under what conditions. strategy → · MVP (Minimum Viable Product)The smallest thing that tests your riskiest assumption: minimum scope to learn, not minimum quality. general → · North Star MetricThe single metric that best captures the value a product delivers to its users. metrics → · NPS (Net Promoter Score)A survey metric measuring willingness to recommend, calculated by subtracting the percentage of Detractors (0-6) from Promoters (9-10) on a 0-10 scale. metrics → · OKRsA goal-setting structure pairing a qualitative objective with two to four measurable key results that confirm the objective was achieved. strategy → · PR/FAQA two-part Amazon document pairing a fictional press release (written as if the product shipped) with FAQs split into external customer questions and hard internal questions about viability, unit economics, and failure risk. general → · PrioritizationThe act of ordering what to build next against a specific outcome, using evidence and judgment, not just a framework score. general → · Product backlogAn ordered list of everything the team might build, used to sequence work against outcomes rather than accumulate requests. general → · Product discoveryThe ongoing work of reducing value, usability, feasibility, and viability risk before committing engineering to build. discovery → · Product discoveryThe continuous practice of identifying which customer problems are worth solving before committing to a solution, running in parallel with delivery rather than as a preceding phase. general → · Product metricsQuantitative signals that tell a product team whether users are getting value and whether the business is capturing it. The foundation of every analytical PM interview question. metrics → · Product requirements document (PRD)A living document that defines the problem, target users, success metrics, scope, and acceptance criteria for a product or feature before engineering begins. general → · product roadmapA prioritized set of bets on which problems to solve, in what order, and under what assumptions. strategy → · Product roadmapA prioritized, time-horizon plan that communicates what a team is building, in what order, and why. A communication artifact as much as a planning tool. general → · Product-led growthA go-to-market strategy where the product itself is the primary channel for acquisition, activation, and expansion. strategy → · Product-market fit (PMF)The state where a product satisfies a real market demand strongly enough that the market pulls it forward without you forcing it. strategy → · Prompt engineeringDesigning and iterating on the instructions given to an LLM to reliably produce outputs that meet quality, format, and cost targets. ai → · PrototypeA throwaway artifact built to answer one specific question about desirability before engineering commits. discovery → · RAG (Retrieval-Augmented Generation)Giving a model relevant documents at query time so its answers are grounded in current, private knowledge. ai → · Retention rateThe share of users from a given cohort who return and perform a meaningful action within a specified time window. metrics → · ScopingThe act of deciding what is in and out of a product or feature before building, and being able to defend those cuts. general → · ScrumA time-boxed iterative framework where cross-functional teams build toward a Sprint Goal every one to four weeks, then inspect the increment and adapt before the next cycle. general → · SprintA fixed time-box of one month or less in which a team produces a usable increment toward a single sprint goal. general → · stakeholderAny person or group whose goals, authority, or resources can materially shape a product decision, and who therefore requires active management by the PM. strategy → · Stakeholder managementA PM's practice of aligning, influencing, and sometimes disappointing the people with a stake in the product, without formal authority over any of them. behavioral → · TAM, SAM, SOMThree concentric market tiers (Total Addressable, Serviceable Addressable, Serviceable Obtainable) that a PM uses to justify roadmap investment, frame market entry, and bound realistic capture. strategy → · Technical debtThe cost of shortcuts taken to ship faster, measured not in code complexity but in future business impact: slowed velocity, inflated costs, and compounding risk. engineering → · ThroughputThe number of work items completed per unit of time, measured as a count of items (not story points). agile → · TokenThe atomic unit an LLM reads, writes, and bills by; roughly 4 characters or 0.75 words in English. ai → · User storyA short, user-centered description of a feature need written in the format "As a [user], I want [action] so that [outcome]," with acceptance criteria defining done. general → · VelocityThe sum of story points for fully completed user stories at the end of a sprint, averaged over 3-5 sprints to forecast future capacity. general → · WSJF (Weighted Shortest Job First)A prioritization formula that ranks work by dividing Cost of Delay by job duration, so the highest-returning item per unit of time is built first. prioritization →