Glossary
Plain-language definitions of the vocabulary modern product and engineering teams use every day — kanban, cycles, backlogs, velocity, SLAs, webhooks, RLS, semantic search, AI agents, propose/approve governance and more.
Clear, tool-agnostic definitions — plus a note on how Planoda puts each idea into practice. Skim a term, or read the full entry.
The glossary in short
Planning & flow
Kanban is a visual workflow method that maps work onto a board of columns representing stages — typically backlog, in progress, and done. Cards move left to right as work advances. It emphasizes continuous flow, making bottlenecks visible, and limiting work in progress rather than committing to fixed time-boxes.
Read definitionA board is a visual layout that arranges work as cards in columns, where each column represents a stage of a workflow — typically backlog, in progress, and done. Cards move left to right as work advances. Boards make a team's flow visible at a glance, exposing where work sits, stalls, or piles up.
Read definitionA cycle — often called a sprint — is a fixed, repeating time-box, usually one or two weeks, during which a team commits to a focused set of work and aims to finish it. Cycles create a regular cadence for planning, focus, and review, turning an open-ended backlog into shippable increments.
Read definitionA sprint is a fixed, repeating time-box — usually one to four weeks — during which a Scrum team commits to a focused set of work and aims to ship a usable increment. It is the Scrum term for the iteration other frameworks call a cycle, giving the team a regular cadence for planning, focus, and review.
Read definitionA backlog is the ordered list of all work a team has identified but not yet started — features, bugs, improvements, and ideas. It is the team's single source of pending work, prioritized so the most valuable or urgent items sit at the top, ready to be pulled into a cycle or onto a board.
Read definitionA WIP (work-in-progress) limit is a cap on how many items may be active in a given workflow stage at once. By forcing a team to finish work before starting more, WIP limits expose bottlenecks, reduce context-switching, and shorten cycle time. They are the core mechanism that makes Kanban flow rather than pile up.
Read definitionA dependency is a relationship where one piece of work cannot proceed or finish until another is done. Dependencies make the order of work matter: a blocked item must wait for its blocker to clear. Tracking them explicitly reveals the true sequence of a project and surfaces the chains that most threaten the timeline.
Read definitionA swimlane is a horizontal row on a board that groups cards by a shared dimension — team, assignee, priority, project, or work type — running across the board's columns. Layered onto a Kanban or sprint board, swimlanes add a second axis of organization, making it easy to see at a glance how work is distributed and to apply policies per lane.
Read definitionTimeboxing is the practice of allocating a fixed, maximum amount of time to an activity in advance, and stopping when that time is spent regardless of completion. It inverts the usual relationship: instead of time expanding to fit the work, the work is constrained to fit the time. The fixed box forces focus, prioritization, and a decision at the boundary.
Read definitionCadence is the regular, predictable rhythm at which a team performs a recurring activity — shipping, planning, reviewing, or releasing. A steady cadence turns events into habits the whole organization can synchronize around. Its value lies in consistency: a reliable beat makes capacity, velocity, and delivery predictable, and lets stakeholders plan around a known schedule.
Read definitionScrum & ceremonies
Agile is a family of software-development approaches built on the 2001 Agile Manifesto, which values working software, customer collaboration, and responding to change over rigid plans and documentation. Rather than one process, it is a set of principles — short iterations, continuous feedback, and incremental delivery — that frameworks like Scrum and Kanban put into practice.
Read definitionScrum is an agile framework for delivering work in fixed iterations called sprints. A small cross-functional team plans a sprint, works without mid-sprint scope changes, then inspects and adapts through review and retrospective. Defined roles, events, and artifacts give the framework structure while leaving the actual engineering practices up to the team.
Read definitionA scrum master is the person accountable for a team's effective use of Scrum. They are a facilitator and coach, not a manager: they run the sprint events, remove impediments blocking the team, shield it from outside disruption, and help the organization adopt agile practices. They have no authority over priorities or who does what work.
Read definitionA product owner is the person accountable for maximizing the value a Scrum team delivers. They own and order the product backlog, decide what gets built and in what sequence, and translate stakeholder needs into clear, prioritized work. They are a single, empowered decision-maker — not a committee — so the team always has one authoritative source of priority.
Read definitionSprint planning is the meeting that opens a sprint, where the team reviews the prioritized backlog and decides what it will commit to in the coming iteration. The team agrees on a sprint goal, pulls a realistic amount of work it believes it can finish, and clarifies scope. The output is a concrete, achievable plan for the time-box ahead.
Read definitionA sprint review is the event at the end of a sprint where the team demonstrates the working increment to stakeholders and gathers feedback. It is a working session, not a status meeting: the team shows what is actually done, stakeholders react, and the product backlog is adjusted in response. The output is an updated plan, not a sign-off.
Read definitionA sprint goal is a single, concise objective the team commits to achieving during a sprint. It states the outcome the sprint is meant to produce, not a checklist of tasks. The goal gives the work coherence: it explains why these items were chosen, guides trade-off decisions mid-sprint, and provides one clear yardstick for whether the sprint succeeded.
Read definitionA standup is a brief, regular team sync — often daily and time-boxed to fifteen minutes — where members share progress, plans, and blockers. Kept short by design (traditionally held standing up), it aligns the team and surfaces obstacles early without becoming a status meeting. The focus is on flow and blockers, not detailed reporting.
Read definitionA retrospective is a recurring meeting, usually at the end of a cycle, where a team reflects on how it worked — what went well, what didn't, and what to change. Its output is a small set of concrete improvements to try next. The retrospective is the engine of continuous improvement, turning experience into deliberate process change.
Read definitionBacklog refinement is the ongoing practice of keeping a product backlog ready to work: clarifying items, splitting large ones, adding estimates and acceptance criteria, and re-ordering by priority. Often called grooming, it is continuous rather than a single event, ensuring the top of the backlog is always well-understood and small enough to be pulled into a sprint.
Read definitionA definition of ready is a shared checklist a backlog item must satisfy before a team accepts it into a sprint. It typically requires that the item is clear, estimated, free of blocking dependencies, and has acceptance criteria. It is the entry gate to work, the mirror image of the definition of done, which is the exit gate.
Read definitionA definition of done is a shared, explicit checklist of what must be true before any work item counts as complete — code reviewed, tests passing, documentation updated, deployed. It removes ambiguity about the word 'done,' preventing half-finished work from being declared finished and creating a consistent quality bar across the whole team.
Read definitionWaterfall is a sequential project management approach where work flows through distinct phases — requirements, design, implementation, testing, deployment — each completed and signed off before the next begins. Plans, scope, and schedule are fixed up front. It suits projects with stable, well-understood requirements but struggles to absorb change discovered mid-project.
Read definitionStories & estimation
A user story is a short, plain-language description of a feature told from the user's perspective, classically in the form 'As a [role], I want [capability], so that [benefit].' It captures who needs something and why, deliberately leaving the how to the team. Stories keep work framed around user value rather than technical tasks.
Read definitionAcceptance criteria are the specific, testable conditions a work item must satisfy to be considered complete and correct. Written before work starts, they define the boundaries of a feature — what it must do, and how you'll know it works — turning a vague request into a checklist everyone agrees on, so 'done' is verifiable rather than a matter of opinion.
Read definitionThe three amigos is a practice where three perspectives — business, development, and testing — review a backlog item together before it is built. The product representative explains the intent, the developer raises feasibility, and the tester probes edge cases. The collaboration surfaces ambiguity and produces shared, testable acceptance criteria early, when gaps are cheapest to fix.
Read definitionStory points are a relative, unitless measure of how much effort an issue will take, accounting for complexity and uncertainty rather than raw hours. Teams estimate in points — often using a Fibonacci-like scale — to compare items against each other quickly. Summed across a cycle, points feed velocity and burndown without false precision about clock time.
Read definitionRelative estimation sizes work by comparing items to each other rather than guessing absolute hours. Teams use techniques like story points, t-shirt sizing (S/M/L), and planning poker to ask 'is this bigger than that?' — a question humans answer more reliably than 'how long will this take?' The sized values then feed velocity and forecasting.
Read definitionPlanning poker is a consensus-based estimation technique where team members privately pick a card representing a story's effort, then reveal simultaneously. Divergent estimates spark discussion, and the team re-votes until it converges. Revealing at once prevents anchoring — no one is swayed by the first or loudest number — so estimates capture the whole team's understanding.
Read definitionT-shirt sizing is a relative estimation technique that classifies work into rough buckets — XS, S, M, L, XL — instead of precise numbers. The coarse scale is fast and intuitive, ideal for early, high-level estimation of epics or roadmap items where detailed point estimates would be false precision. It compares relative size rather than predicting exact effort.
Read definitionA spike is a time-boxed piece of investigative work used to reduce uncertainty before committing to a feature. Instead of producing shippable functionality, it produces knowledge — a prototype, a technical answer, or a risk assessment. The output is what the team learns, and the strict time-box prevents open-ended research from consuming a sprint.
Read definitionStory mapping is a technique for organizing user stories into a two-dimensional model of a product. The horizontal axis is the user's journey, step by step; the vertical axis ranks the stories under each step by priority. Slicing horizontally produces coherent releases, making it a powerful tool for planning a minimum viable product and sequencing releases.
Read definitionA work breakdown structure (WBS) is a hierarchical decomposition of a project into progressively smaller, deliverable-oriented pieces. The top is the overall outcome; each level below divides it into more concrete components, down to tasks small enough to estimate and assign. It organizes total scope so nothing is forgotten and effort can be rolled up reliably.
Read definitionCapacity planning is the practice of matching a team's committed work to the effort it can realistically deliver in a given period. It accounts for headcount, availability, holidays, meetings, and support load — not just an idealized full-time team. Planning to real capacity, rather than maximum capacity, keeps commitments achievable and prevents chronic over-loading of a sprint.
Read definitionStrategy & goals
A roadmap is a high-level, time-oriented view of what a team or product plans to build and roughly when. It communicates direction and sequencing across initiatives and projects, aligning stakeholders on priorities. Unlike a backlog of granular tasks, a roadmap operates at the altitude of themes, outcomes, and quarters rather than individual issues.
Read definitionAn initiative is a large, strategic body of work that spans multiple projects, teams, or cycles toward a single outcome. It sits above projects and issues in the planning hierarchy, grouping related efforts under one goal. Initiatives let leadership track progress on big bets without drowning in individual tickets.
Read definitionA project is a bounded body of work with a defined goal, scope, and usually an end date — a feature launch, a migration, a redesign. It groups the related issues that deliver that outcome, sitting above individual issues and below strategic initiatives in the planning hierarchy, so a team can track one cohesive effort as a unit.
Read definitionAn epic is a large unit of work too big to finish in a single cycle, broken down into smaller related issues that ship incrementally. It groups those child issues under one theme and tracks their combined progress. Epics sit between individual issues and broader projects or initiatives in the planning hierarchy.
Read definitionA theme is a high-level area of focus that groups related work under a common strategic intent. Broader than an epic, a theme spans multiple epics or initiatives and often persists across several quarters. Themes give a roadmap coherence by answering not just what a team is building but the larger goal all that work collectively serves.
Read definitionAn issue tracker is the system of record for a team's work — every bug, feature, and task captured as a structured issue with a state, assignee, priority, and history. It replaces scattered emails and spreadsheets with one searchable, accountable source of truth that the whole team plans, executes, and reports against.
Read definitionAn SLA (service-level agreement) is a defined commitment to respond to or resolve certain work within a set time — for example, acknowledging urgent bugs within an hour. SLAs turn vague expectations into measurable targets, hold teams accountable, and signal which work the clock is running on so the most time-sensitive items aren't lost.
Read definitionTriage is the process of reviewing newly reported issues and deciding what happens to each — accept and prioritize, request more detail, route to a team, or close. Borrowed from emergency medicine, it keeps the incoming flow of bugs and requests from overwhelming a team by quickly sorting signal from noise at the front door.
Read definitionOKR is a goal-setting framework that pairs a qualitative Objective — what you want to achieve — with three to five measurable Key Results that prove you got there. Set per quarter and scored at the end, OKRs align a team on a small number of outcomes, keeping effort focused on results rather than a list of activities.
Read definitionA North Star metric is the single measure that best captures the core value a product delivers to customers — and that, when it grows, reliably pulls revenue and retention up with it. It aligns an entire company on one number, cutting through competing departmental metrics so every team can see how its work moves the thing that matters most.
Read definitionA PRD (product requirements document) defines what a product or feature should do and why, before it is built. It captures the problem, the target users, the goals and success metrics, the scope, and the requirements — giving design, engineering, and stakeholders one shared reference so everyone builds toward the same outcome rather than their own interpretation.
Read definitionA milestone is a significant checkpoint in a project — a meaningful date or deliverable that marks progress, such as a beta launch, a feature freeze, or a public release. Unlike a task, a milestone has no duration; it is a moment that signals a phase is complete, used to coordinate teams and communicate timing to stakeholders.
Read definitionScope creep is the gradual, uncontrolled expansion of a project's requirements after work has begun — small additions and 'while we're at it' requests that accumulate until the original plan no longer fits the time or budget. It is a leading cause of missed deadlines, because each addition feels minor but the total quietly overruns the commitment.
Read definitionA Gantt chart is a horizontal bar chart that lays project tasks along a timeline, with each bar's position and length showing when a task starts, how long it lasts, and how it overlaps with others. Dependency lines connect tasks that must happen in order. It is the classic view for planning sequence and seeing a schedule at a glance.
Read definitionThe critical path is the longest chain of dependent tasks in a project — the sequence that determines the shortest possible completion time. Any delay to a task on the critical path delays the whole project; tasks off it have slack. Identifying the critical path tells a team exactly where timing is fragile and where to focus to protect the deadline.
Read definitionTime to market (TTM) is the elapsed time from when a product or feature is first conceived until it is available to customers. It measures how quickly an organization can turn an idea into something usable. A shorter time to market lets a team capture opportunity sooner, learn from real users faster, and respond to competitors before the window closes.
Read definitionAn objective is the qualitative, time-boxed outcome a team commits to in an OKR — the memorable 'what we're trying to achieve this quarter.' It sets direction; its key results make it measurable. An objective without key results is a slogan; key results without an objective are a dashboard.
Read definitionA key result is the measurable half of an OKR — a specific metric with a starting value, a target, and a current value that proves whether an objective is being achieved. Where the objective is qualitative ('make onboarding effortless'), the key result is quantitative ('activation rate from 38% to 60%').
Read definitionDiscovery & prioritization
Jobs to Be Done is a framework for understanding why customers adopt a product: people "hire" products to make progress on a specific job in a given circumstance. It shifts focus from customer demographics and product features to the underlying goal — the job — revealing the real competition and the true criteria by which customers judge success.
Read definitionProduct-market fit is the point at which a product satisfies a strong market demand — the right product serving the right market so well that growth begins to pull rather than push. It is the milestone before which a startup should focus on finding fit, and after which it should focus on scaling, often felt as demand outrunning the team's ability to keep up.
Read definitionTime to value (TTV) is the elapsed time between a user starting with a product and reaching their first meaningful outcome — the moment the product's promise becomes real to them. Shorter TTV strongly predicts activation and retention, so reducing it, often by engineering a fast "aha moment," is a central goal of onboarding and product design.
Read definitionRICE is a prioritization framework that scores each initiative by Reach (how many people it affects), Impact (how much it moves the needle per person), Confidence (how sure the estimates are), and Effort (the work required). The score is Reach × Impact × Confidence ÷ Effort, producing a comparable number that ranks competing ideas by expected value per unit of work.
Read definitionA RACI matrix is a responsibility-assignment chart that clarifies who does what on a project by labeling each person's role for each task as Responsible (does the work), Accountable (owns the outcome, one per task), Consulted (gives input), or Informed (kept updated). It removes ambiguity over ownership and decision rights before work begins.
Read definitionThe Eisenhower Matrix is a 2×2 prioritization grid that sorts tasks by two axes — urgent versus not urgent, and important versus not important — producing four quadrants, each with a prescribed action: do it now, schedule it, delegate it, or eliminate it. It helps people separate genuine priorities from work that merely feels pressing.
Read definitionMoSCoW is a prioritization method that sorts requirements into four categories — Must have, Should have, Could have, and Won't have (this time). The capitalized letters form the name; the lowercase o's make it pronounceable. It forces a team to agree explicitly on what is essential versus deferrable, rather than treating every request as equally urgent.
Read definitionICE is a lightweight prioritization framework that scores each idea on three factors — Impact (how much it will move the goal), Confidence (how sure you are in the estimate), and Ease (how simple it is to implement) — usually on a 1–10 scale. The ICE score is Impact × Confidence × Ease, giving a fast, comparable number for ranking competing experiments.
Read definitionWSJF is a prioritization model from SAFe that sequences work for maximum economic benefit by dividing the Cost of Delay by the job's duration or size. The shortest jobs with the highest cost of delay rank first. Cost of Delay itself sums user/business value, time criticality, and risk reduction or opportunity enablement, each scored on a relative scale.
Read definitionThe Kano model classifies product features by how their presence or absence affects customer satisfaction, sorting them into five types: Must-be (basics expected by default), Performance (more is better), Attractive (delighters), Indifferent (no effect either way), and Reverse (please some, annoy others). Its core insight is that satisfaction is asymmetric — a missing basic angers users while its presence earns no credit.
Read definitionA minimum viable product is the smallest version of a product that can be released to learn the most about customers with the least effort. Rather than building everything, a team ships just enough to test whether the core idea solves a real problem, then uses real-world feedback to decide what to build, change, or abandon next.
Read definitionThe Lean Startup is a methodology, formalized by Eric Ries, for building products under extreme uncertainty by treating ideas as hypotheses to be tested cheaply. Teams cycle through a build-measure-learn loop — ship a minimum viable product, measure real behavior, learn, then persevere or pivot — to find a sustainable business with minimal wasted effort.
Read definitionContinuous discovery is the practice of engaging with customers regularly — ideally weekly — to inform product decisions, rather than running discovery as a one-off phase before delivery. Small cross-functional teams interview users, spot unmet needs, test solution ideas, and feed that learning directly into what they build, keeping discovery and delivery running in parallel.
Read definitionAn opportunity solution tree is a visual map, created by Teresa Torres, that connects a desired outcome at the top to the customer opportunities (unmet needs and pain points) beneath it, then to candidate solutions and the assumption tests that validate them. It keeps a team focused on outcomes and makes the reasoning behind product decisions explicit.
Read definitionA customer journey map is a visual representation of every step a customer takes to accomplish a goal with a product or company — across stages, touchpoints, channels, and over time. It captures what the customer does, thinks, and feels at each step, exposing friction, gaps, and moments of delight so teams can improve the end-to-end experience.
Read definitionFlow metrics & forecasting
A burndown chart tracks remaining work against time over a cycle, sloping from the total scope down toward zero as items are completed. It shows whether a team is on pace to finish what it committed to, making slippage visible early. The ideal line falls steadily; a flat line warns that work is stalling.
Read definitionA burnup chart tracks completed work rising toward a total-scope line over time. Unlike a burndown, which only shows remaining work falling toward zero, a burnup plots two lines — work done and total scope — so scope changes are visible as movement in the upper line. It distinguishes 'falling behind' from 'scope was added.'
Read definitionVelocity is the average amount of work a team completes per cycle, measured in issues or story points. By tracking it over several cycles, teams forecast how much they can realistically take on next. Velocity is a planning aid for a specific team over time — never a target to maximize or a way to compare teams against each other.
Read definitionThroughput is the number of work items a team completes in a given period — issues finished per week, for example. It is the simplest flow metric: a direct count of output over time. Tracked across periods, throughput reveals a team's real delivery capacity and is the basis for probabilistic, estimate-free forecasting.
Read definitionCycle time is how long an issue takes from the moment work actively starts on it to the moment it is done. Measured in hours or days, it captures the team's hands-on flow efficiency. Shorter, more consistent cycle times mean a more predictable system — the core flow metric Kanban teams optimize.
Read definitionLead time is the total elapsed time from when an issue is first created or requested to when it is delivered. Unlike cycle time, it includes the waiting period in the backlog before work begins. Lead time reflects the customer's experience of how long a request actually takes end to end.
Read definitionFlow efficiency is the percentage of an item's total time that was spent actively worked on versus waiting. If an issue takes ten days from start to done but only two were hands-on, its flow efficiency is 20%. Most teams are shocked to find theirs is low — proof that delays, not slow work, dominate delivery time.
Read definitionA cumulative flow diagram (CFD) is a stacked area chart showing how many items sit in each workflow state over time. The colored bands reveal flow health at a glance: a widening band means work is piling up in that stage, a steady set of parallel bands means smooth flow. It is Kanban's richest single picture of a system's behavior.
Read definitionLittle's Law is a foundational result from queueing theory stating that, on average, the number of items in a system equals the rate items arrive (or complete) multiplied by the average time each spends in the system. For teams: average work in progress equals throughput times cycle time — which is why limiting WIP directly shortens cycle time.
Read definitionValue stream mapping is a lean technique for diagramming every step work passes through from request to delivery, measuring the time spent in each. By distinguishing value-adding time from waiting and rework, it exposes where flow stalls. The goal is to maximize flow efficiency — the share of total lead time during which work is actively progressing.
Read definitionDORA metrics are four research-backed measures of software delivery performance: deployment frequency, lead time for changes, change failure rate, and time to restore service. Identified by the DevOps Research and Assessment program, they balance speed (the first two) against stability (the last two), giving engineering teams an evidence-based scorecard for how well they ship.
Read definitionMonte Carlo forecasting predicts delivery by running thousands of simulated futures from a team's historical throughput, producing a probability rather than a single date. Instead of 'we'll finish on the 20th,' it answers 'an 85% chance of finishing by the 24th.' It replaces false-precision estimates with honest, evidence-based ranges drawn from how the team actually performs.
Read definitionA least-squares forecast fits a straight line to a series of data points — minimizing the sum of squared distances from the line — and extrapolates it to project a future value. In delivery forecasting it's the simplest honest projection: fit a trend to the remaining-scope burndown and read off where the line crosses zero to estimate a completion date.
Read definitionA delivery forecast projects when work will actually finish, based on the team's observed throughput rather than a wished-for deadline. Instead of asking everyone to re-estimate, it fits a trend to the remaining-scope trajectory and extrapolates the completion date — and says so honestly when the data won't support a projection.
Read definitionDelivery forecasting is projecting when work will finish from a team's observed throughput rather than from a chosen deadline. It fits a trend to the remaining-scope trajectory and extrapolates the completion date — and, in its probabilistic form, returns a range with confidence levels instead of a single date, so commitments are grounded in evidence rather than optimism.
Read definitionProduct & growth metrics
Churn rate is the percentage of customers (or revenue) lost over a period, calculated as customers lost divided by customers at the start of the period. It is the inverse of retention and the single most-watched health metric for subscription businesses, because small monthly losses compound into large annual ones.
Read definitionRetention rate is the percentage of customers (or users) who remain active over a period — the mirror image of churn. Calculated as customers retained divided by customers at the period's start, it measures whether a product delivers durable, repeated value rather than a one-time hit, and underpins almost every other growth metric.
Read definitionCohort analysis groups users by a shared starting characteristic — most often their signup or first-purchase date — and tracks how each group behaves over time. By comparing cohorts side by side, it separates the effect of when someone joined from the effect of how long they have been around, revealing trends that blended averages hide.
Read definitionActivation rate is the percentage of new users who reach a defined first-value milestone — the moment they experience the product's core benefit. Sitting between signup and retention in the funnel, it measures whether onboarding actually delivers on the promise that brought users in, and is one of the strongest early predictors of whether they will stay.
Read definitionThe DAU/MAU ratio divides daily active users by monthly active users to gauge engagement intensity — the share of a product's monthly audience that returns on a typical day. Expressed as a percentage, it is a stickiness measure: a ratio of 50% implies the average monthly user is active about fifteen days a month.
Read definitionNet Promoter Score measures customer loyalty from a single question: how likely are you to recommend us, on a 0–10 scale. Respondents scoring 9–10 are promoters, 7–8 passives, and 0–6 detractors. NPS is the percentage of promoters minus the percentage of detractors, yielding a number from −100 to +100.
Read definitionCustomer Satisfaction Score (CSAT) measures how satisfied customers are with a specific interaction, product, or service, usually by asking them to rate it on a short scale. CSAT is the percentage of respondents who chose a satisfied rating (commonly the top one or two of a five-point scale), making it a direct, transactional gauge of a single moment.
Read definitionA customer health score is a composite metric that blends behavioral and relationship signals — product usage, support history, engagement, sentiment, and payment status — into a single indicator of how likely an account is to renew, expand, or churn. It gives customer success teams an early, prioritized view of which accounts need attention.
Read definitionAnnual Recurring Revenue (ARR) is the normalized, predictable subscription revenue a business expects over a year, counting only recurring contracts and excluding one-time fees. It is the headline scale metric for subscription companies — a snapshot of the run-rate revenue the customer base would generate over twelve months at the current moment.
Read definitionCustomer Acquisition Cost (CAC) is the total sales and marketing spend required to win one new customer, calculated as those fully loaded costs over a period divided by the new customers acquired in it. CAC is half of the core unit-economics equation: a business is viable only when the value a customer returns comfortably exceeds the cost to acquire them.
Read definitionCustomer Lifetime Value (LTV or CLV) is the total profit a business expects to earn from a customer across the entire relationship. A common estimate is average revenue per customer times gross margin, divided by churn rate. LTV quantifies what a customer is truly worth, setting the ceiling on what a business can sensibly spend to acquire and keep them.
Read definitionConversion rate is the percentage of people who complete a desired action out of those who had the opportunity — visitors who sign up, trials that become paid, or leads that close. Calculated as conversions divided by the eligible population, it is the fundamental efficiency measure of any funnel step, isolating how well one transition performs.
Read definitionFunnel analysis tracks how users move through a sequence of steps toward a goal — such as visit, signup, activate, purchase — measuring the conversion rate and drop-off at each stage. By revealing exactly where the most users leak out, it directs improvement effort to the single step where fixing it yields the greatest gain.
Read definitionA/B testing is a controlled experiment that randomly splits users into a control group and one or more variants, then compares a target metric to learn which version performs better. By isolating a single change and relying on randomization, it establishes a causal link between that change and the outcome rather than mere correlation.
Read definitionA leading indicator is a metric that predicts a future outcome — it moves before the result it foreshadows, giving teams time to act. Activation rate, trial signups, and pipeline coverage are leading indicators of revenue. Because they shift early, they are levers teams can influence now, unlike outcomes that are already settled by the time they appear.
Read definitionA lagging indicator is a metric that confirms an outcome after it has occurred — it reflects results already produced rather than predicting them. Revenue, churn rate, and customer lifetime value are lagging indicators: trustworthy and unambiguous, but slow to respond, so by the time they move the underlying cause is already in the past.
Read definitionA vanity metric is a number that looks impressive but does not inform decisions or correlate with real success — total registered users, page views, or app downloads. It tends to only go up, lacks context for action, and flatters rather than informs, making it a poor basis for strategy compared to actionable, comparable metrics.
Read definitionPirate Metrics, or AARRR, is a framework that organizes a product's growth into five stages: Acquisition, Activation, Retention, Referral, and Revenue. Named for the sound of its initials, it gives teams a shared map of the customer lifecycle so each stage gets its own metric, owner, and improvement effort rather than being lumped into one vague 'growth' goal.
Read definitionDelivery & releases
CI/CD is the practice of automatically building, testing, and releasing code through a pipeline triggered by every change. Continuous integration merges and verifies work frequently to catch conflicts early; continuous delivery keeps the software always in a releasable state, often deploying automatically. Together they shorten the path from a commit to running in production.
Read definitionA pull request (PR), also called a merge request, is a proposal to merge a set of code changes from one branch into another. It packages a diff with a description, opens it for review and discussion, runs automated checks, and serves as the gate where changes are inspected and approved before they enter the main codebase.
Read definitionCode review is the practice of having one or more engineers examine a change before it merges, checking for correctness, clarity, security, and adherence to conventions. Beyond catching defects, it spreads knowledge across the team, enforces shared standards, and creates a documented rationale for why code looks the way it does — typically conducted on a pull request.
Read definitionA feature flag (or feature toggle) is a runtime switch that turns a piece of functionality on or off without deploying new code. It decouples deployment from release, letting teams ship code dark, roll a feature out gradually, target specific users, run experiments, and instantly disable a problem — all through configuration rather than a redeploy.
Read definitionA canary release is a deployment strategy that rolls new code out to a small subset of users or servers first, monitors it closely for errors or regressions, and only proceeds to a full rollout if it stays healthy. Named after the canary in a coal mine, it limits the blast radius of a bad change and enables fast, low-risk rollback.
Read definitionBlue-green deployment runs two identical production environments — one live (blue), one idle (green). You deploy the new release to the idle environment, verify it, then switch traffic over in a single cut. If anything goes wrong, you flip traffic back instantly. It enables near-zero-downtime releases and an immediate rollback path.
Read definitionA rollback is reverting a system to a previous known-good state after a release introduces a problem. It is the fastest way to restore service when a deploy goes wrong: rather than diagnosing and forward-fixing under pressure, you return to the version that worked. A reliable, fast rollback path is a hallmark of mature deployment practice.
Read definitionTechnical debt is the implied future cost of choosing an easier or faster solution now over a better one that takes longer. Like financial debt, it accrues interest: shortcuts in code, architecture, or tests slow every future change until they're repaid through refactoring. Some debt is deliberate and strategic; some is accidental and corrosive.
Read definitionA blocker is anything that prevents an issue from progressing until it is resolved — a dependency on unfinished work, a pending decision, a missing access, or an external delay. Marking work as blocked makes hidden stalls visible so the team can clear them deliberately, rather than letting issues quietly age in progress.
Read definitionInfrastructure as code is the practice of defining and provisioning servers, networks, databases, and other infrastructure through machine-readable configuration files rather than manual setup. Because the definition lives in version control, environments become reproducible, reviewable, and auditable — you can recreate or change infrastructure by editing code rather than clicking through a console.
Read definitionSemantic versioning is a convention for numbering software releases as MAJOR.MINOR.PATCH, where each part signals the kind of change. You increment MAJOR for incompatible (breaking) changes, MINOR for backward-compatible new features, and PATCH for backward-compatible bug fixes. The version number itself communicates how risky an upgrade is, so dependents can update safely.
Read definitionPair programming is a software development technique where two engineers work together at one workstation: one (the driver) writes the code while the other (the navigator) reviews each line, thinks ahead, and catches issues in real time. The pair swaps roles frequently. It trades some short-term throughput for higher quality, faster learning, and shared knowledge.
Read definitionTest-driven development is a practice where you write an automated test for a behavior before writing the code that makes it pass. It follows a tight red-green-refactor loop: write a failing test, write the minimum code to pass it, then clean up the design. Tests drive the implementation, producing well-covered, intentionally designed code.
Read definitionReliability & operations
Incident management is the coordinated process of detecting, responding to, and resolving unplanned disruptions to a service, then restoring normal operation as fast as possible. It defines roles (incident commander, communications lead), severity levels, escalation paths, and a status-communication cadence, with the goal of minimizing impact and learning from every failure.
Read definitionA postmortem is a structured, blameless write-up produced after an incident that documents what happened, the timeline, the contributing factors, and the actions taken, then defines concrete follow-ups to prevent recurrence. Its purpose is organizational learning, not punishment — it treats failures as systemic, examining process and tooling rather than assigning individual blame.
Read definitionA service-level objective (SLO) is a measurable target for a system's reliability over a window — for example, 99.9% of requests succeeding in 30 days. It is set against a service-level indicator (a metric like success rate or latency) and is the internal goal that informs the externally promised SLA, giving teams a precise definition of "reliable enough."
Read definitionAn error budget is the amount of unreliability a team is allowed to spend, derived as the inverse of a service-level objective — a 99.9% SLO permits 0.1% failure. It reframes reliability as a finite resource: when budget remains, teams ship boldly; when it's exhausted, they pause risky changes and prioritize stability until it replenishes.
Read definitionObservability is the degree to which you can understand a system's internal state from the data it emits — logs, metrics, and traces. A system is observable when you can answer new questions about its behavior without shipping new code, letting you debug unknown failures in production rather than only the ones you anticipated.
Read definitionDistributed tracing follows a single request as it travels across the many services that handle it, stitching each hop into one end-to-end timeline. Each unit of work is a span; spans share a trace ID and nest into a tree. Tracing reveals where time is spent and which service failed in a system too distributed to debug from logs alone.
Read definitionTelemetry is the automatic collection and transmission of data about how a system behaves — performance metrics, error events, usage signals, and traces — from where it runs to where it can be analyzed. It is the raw material of observability: without telemetry flowing out of a system, you cannot see inside it or reason about its health.
Read definitionRate limiting caps how many requests a client may make to a service within a time window — for example 100 requests per minute per API key. It protects a system from being overwhelmed by accidental loops, abusive traffic, or noisy neighbors, and enforces fair usage across tenants. Clients over the limit receive a 429 response, often with a retry hint.
Read definitionA circuit breaker is a resilience pattern that stops an application from repeatedly calling a failing dependency. After errors cross a threshold, the breaker trips open and fails calls fast instead of waiting on timeouts, giving the struggling service room to recover. It periodically tries again, closing once the dependency is healthy — preventing one failure from cascading into a system-wide outage.
Read definitionGraceful degradation is designing a system so that when a component fails, the system continues operating with reduced functionality rather than collapsing entirely. A non-essential feature going down should produce a smaller, contained loss — a disabled widget or a cached fallback — not a total outage. It keeps the core experience available even as parts of the system fail.
Read definitionMean time to recovery is the average time it takes to restore service after a failure or incident — measured from when the failure begins until normal operation resumes. One of the four DORA metrics, it gauges how quickly a team detects, diagnoses, and recovers from problems. A low MTTR signals resilience: failures happen, but they're contained fast.
Read definitionAnomaly detection is automatically flagging data points that deviate meaningfully from an expected pattern — a sudden spike in bug inflow, a cycle whose throughput collapses, an SLA breaching its trend. Instead of waiting for someone to notice on a dashboard, the system continuously compares live signals against a baseline and surfaces the outliers worth a human's attention.
Read definitionDrift detection is monitoring for slow, cumulative divergence between two things that should stay aligned — a model's behavior versus its training distribution, a deployed schema versus the code that defines it, a process versus its intended baseline. Unlike anomaly detection's sudden spikes, drift is gradual, so it's caught by comparing state against a reference rather than by watching for a jump.
Read definitionAI & agents
An AI agent is a software system that uses a large language model to pursue a goal across multiple steps — reading context, choosing tools, and taking actions — rather than answering a single prompt. In a work platform, agents triage issues, draft updates, and execute multi-step tasks as autonomous teammates, bounded by the permissions and approvals their operators set.
Read definitionAn agentic workflow is a process in which one or more AI agents carry out a multi-step task with some autonomy — planning, calling tools, and acting on results in a loop — rather than a human driving each step. The agent decides the next action toward a goal, within boundaries its operator sets, turning AI from a single-response assistant into a worker.
Read definitionPropose/approve is a governance pattern for autonomous software: instead of executing a consequential action directly, an AI agent emits it as a proposal that a human or policy must approve before it runs. It keeps fast, read-only work autonomous while gating destructive or irreversible operations — the practical way to give agents real power without surrendering control.
Read definitionA large language model is an AI system trained on vast amounts of text to predict and generate language, enabling it to answer questions, summarize, write, and reason over natural language. LLMs power modern AI assistants and agents. They are probabilistic next-token predictors, which makes them remarkably capable but also prone to confident errors.
Read definitionRetrieval-augmented generation (RAG) grounds a language model's answers in your own data by retrieving the most relevant documents at query time and feeding them into the model's context before it responds. Instead of relying solely on what the model memorized in training, RAG lets it answer from current, authoritative, private sources — sharply reducing hallucination.
Read definitionA context window is the maximum amount of text — measured in tokens — that a language model can consider at once, spanning its input and its output. Everything the model 'sees' for a given request must fit inside it: the prompt, retrieved data, conversation history, and the answer. When content exceeds the window, the oldest or least relevant parts must be dropped.
Read definitionAn AI hallucination is output from a language model that is fluent and confident but factually wrong, fabricated, or unsupported by the model's inputs — invented citations, fake quotes, or made-up details. It arises because models predict plausible text rather than retrieve verified facts, making hallucination an inherent risk that must be mitigated, not a bug that's simply fixed.
Read definitionPrompt injection is an attack on AI systems where malicious instructions hidden in user input or retrieved content trick a language model into ignoring its original directives — leaking data, performing unauthorized actions, or producing harmful output. Because models can't reliably separate trusted instructions from untrusted data in their context, it is a fundamental security risk for any LLM application.
Read definitionSemantic search finds results by meaning rather than exact keywords, using vector embeddings that place similar concepts near each other in mathematical space. A query for 'login broken' can surface an issue titled 'users can't authenticate' even with no shared words. It powers more relevant search and is the retrieval layer behind many AI features.
Read definitionAn MCP server implements the Model Context Protocol, an open standard that lets AI assistants connect to external tools and data through a uniform interface. Rather than building a bespoke integration per assistant, a tool exposes one MCP server describing its available actions and resources, and any MCP-capable AI can discover and use them safely.
Read definitionFine-tuning is the process of further training a pretrained language model on a smaller, task-specific dataset so it adapts to a particular domain, format, or style. Rather than training from scratch, it adjusts the existing model's weights on curated examples, producing a specialized variant that performs a narrow task more reliably than the base model.
Read definitionFew-shot prompting is the technique of including a handful of worked examples in a prompt to show a model exactly what you want before asking it to perform the task. The examples demonstrate the input-output pattern, format, and style, steering the model far more reliably than instructions alone — without any retraining of the underlying model.
Read definitionZero-shot prompting is asking a model to perform a task using instructions alone, with no examples provided. It relies on the broad capabilities the model learned in training to generalize to a task it was never explicitly shown. It is the simplest, fastest way to prompt, and modern models handle many tasks well from a clear instruction by itself.
Read definitionChain-of-thought is a prompting technique that asks a model to reason through a problem step by step before giving its final answer. By generating intermediate reasoning rather than jumping straight to a conclusion, models perform markedly better on multi-step tasks like math, logic, and complex analysis — the explicit steps act as scratch work that improves the result.
Read definitionA system prompt is a set of instructions given to a language model before any user input, defining its role, behavior, tone, constraints, and the tools or knowledge it should use. It frames every response in the conversation that follows. Unlike a user message, it sets the persistent ground rules the model is meant to honor throughout an interaction.
Read definitionTemperature is a setting that controls how random a language model's output is. At low temperature the model strongly favors its most likely next token, producing focused, deterministic-leaning text; at high temperature it samples more broadly, producing varied, creative, and less predictable output. It tunes the trade-off between consistency and diversity in generated responses.
Read definitionA token is the unit a language model reads and generates — typically a word fragment of a few characters rather than a whole word. Text is split into tokens before processing, and models bill, limit, and reason in tokens, not characters. As a rough guide, one token is about four characters or three-quarters of a word in English.
Read definitionInference is the act of running a trained model to produce output from new input — the moment a model is actually used, as opposed to trained. Every prompt sent to a language model triggers an inference pass. Inference dominates the running cost and latency of AI features, since it happens on every request while training happens once.
Read definitionAgent memory is how an AI agent retains and reuses information across turns and sessions, beyond what fits in a single context window. It spans short-term working memory within a task and long-term memory persisted in a store and recalled later, letting an agent remember past interactions, decisions, and learned facts instead of starting fresh every time.
Read definitionTool calling is the capability that lets a language model invoke external functions — query a database, search the web, create a record — by emitting a structured request the surrounding system executes. Instead of only producing text, the model can take actions and pull in live data, turning it from a text generator into the reasoning core of an agent.
Read definitionGuardrails are the controls that constrain what an AI system is allowed to do or say, enforced around the model rather than left to the model's judgment. They validate inputs and outputs, block unsafe or off-policy actions, scope which tools an agent may use, and route risky operations through approval — turning a probabilistic model into a system safe to deploy.
Read definitionModel evaluation is the practice of measuring how well an AI model or system performs against defined criteria, using test cases, scoring methods, and metrics. It replaces 'it seems to work' with evidence — catching regressions, comparing models, and verifying quality before and after deployment. For generative systems, evaluation is harder because outputs rarely have one correct answer.
Read definitionGrounding is the practice of tying an AI model's responses to verifiable, external source data rather than relying solely on what it absorbed during training. By supplying relevant, current, authoritative context at query time — and ideally citing it — grounding reduces hallucination and keeps answers accurate, traceable, and specific to the user's actual data.
Read definitionA multimodal model can process and reason across more than one type of input — combining text with images, audio, video, or other data in a single model. Instead of handling each modality with a separate system, it understands them together, so it can describe a screenshot, answer questions about a chart, or relate a diagram to its caption.
Read definitionA model gateway is a unified layer that sits between an application and one or more AI model providers, routing requests through a single interface. It centralizes provider abstraction, authentication, rate limiting, cost tracking, caching, and failover — so an app can swap models, enforce spending budgets, and observe usage without rewiring every call site.
Read definitionA coding agent is an AI agent that takes a tracked issue all the way to a reviewed pull request — reading the codebase, planning a change, writing and running tests, and opening a PR a human merges. It is distinct from inline code completion: it owns a multi-step task end to end, under review, rather than suggesting the next line.
Read definitionAgent orchestration is the coordination of multiple AI agents — and the hand-offs between them — to complete work no single agent owns end to end. A planner decomposes a goal, specialist agents execute sub-tasks, and results are routed and merged. The hard part is not running the agents but governing what each is allowed to do as they hand work back and forth.
Read definitionAn AI notetaker is an assistant that joins a meeting — a call or a recorded room — transcribes it, and produces a structured summary with decisions and action items. The 2026 version goes further: it turns those action items into tracked work, creating assigned tasks with owners and due dates directly in the team's tool instead of leaving them in a document.
Read definitionAn assignable agent is an AI agent modeled as a first-class workspace member: it has an identity you can @mention, assign issues to, and schedule, just like a human teammate. Instead of living in a separate chat panel, it appears in the assignee picker and the activity feed — so delegating to AI uses the same gestures as delegating to a person.
Read definitionA natural-language dashboard is one you build by describing it in plain language — 'show open bugs by team this cycle, with a burndown' — and the system generates the validated widgets and queries for you. Instead of dragging chart components and writing filters by hand, you state the question; the tool composes the layout from your real data.
Read definitionAn agent skill is a reusable, named agent capability: a packaged workflow with a description, a system prompt, an allowed set of tools, and triggers that decide when it runs. Instead of re-explaining a task every time, you define the skill once and the agent applies it on demand — turning ad-hoc prompting into a library of repeatable, governed capabilities.
Read definitionAn AI cost ledger is a transparent, per-workspace record of metered AI spend: every model call's cost is attributed and accumulated so a team can see exactly what its AI usage costs in real money. It is the alternative to opaque 'credits,' which abstract spend into a proprietary unit whose dollar value and burn rate are hard to reason about.
Read definitionMetered AI credits are consumption-based pricing for AI features: on top of a per-seat fee, a team buys or is granted a pool of credits that each AI action draws down. It became the dominant 2026 pattern because model inference has a real marginal cost. The tradeoff is that credits, denominated in a vendor unit, make spend harder to predict than a flat fee.
Read definitionAn autonomous agent is an AI system that runs its own loop toward a goal — observing, deciding, acting, and re-evaluating — without a human approving each step. Autonomy sits on a spectrum from suggest-only to fully unsupervised; the practical question is not whether an agent is autonomous but how much reach it has and what guardrails bound the actions it can take alone.
Read definitionAgent assignment is handing a unit of work to an AI agent using the same gesture you'd use for a person — setting it as an issue's assignee. The agent then picks up the work, acts on it, and reports back in the activity feed. It makes delegating to AI a first-class part of the workflow instead of a separate chat, while keeping the agent's consequential actions under governance.
Read definitionAn agent session is a single bounded run of an AI agent toward a goal — its sequence of observations, tool calls, and results, with an attached identity, capability scope, budget, and audit record. Treating a run as a first-class session is what makes an agent's behavior governable and reviewable: it has a beginning, an accountable actor, and a transcript you can later distill or inspect.
Read definitionAn agent proposal is a consequential action an AI agent wants to take, packaged as a described, reviewable intent that waits for approval before it executes — rather than running on the model's own decision. It is the concrete unit of propose/approve governance: a record of exactly what the agent intends, surfaced to a human or policy that grants or denies it.
Read definitionThe Model Context Protocol (MCP) is an open standard for connecting AI assistants to external tools and data through one uniform interface. A system exposes its capabilities once as an MCP server — typed tools, readable resources, prompts — and any MCP-capable client can discover and invoke them, replacing a bespoke integration per assistant with a write-once contract.
Read definitionPrompt grounding is supplying a model with the specific, retrieved facts it needs at generation time — pulled from your real data — so its answer is anchored to evidence rather than to its parametric memory. It is the mechanism behind retrieval-augmented generation: ground the prompt in fetched context, and the model cites what's true for you instead of confabulating a plausible-sounding answer.
Read definitionAn AI budget gate is a spend control that checks an agent operation against a workspace's remaining AI budget before it runs and records the actual cost after — so AI usage can never silently exceed its bounds. It turns metered AI from an open-ended liability into a governed resource, the spend half of agent governance alongside the action-approval half.
Read definitionAI triage is using a model to sort incoming work — bugs, support requests, inbound issues — by reading each item and proposing a priority, assignee, labels, and routing. Instead of a human eyeballing a queue, the agent applies consistent rules at scale and surfaces its reasoning, while a person retains the final say on anything consequential.
Read definitionAn AI work platform is a unified system for planning and tracking work where AI is native to every surface rather than bolted on as a side panel. It collapses the categories of issue tracker, docs, projects, and dashboards onto one data model so an agent can reason and act across all of them — the 2026 successor to single-purpose tools stitched together by integrations.
Read definitionAgent governance is the set of controls that make an AI agent's actions safe, attributable, and reviewable: human approval gates on consequential actions, an immutable audit trail of who approved what, role-based capability limits, and spend controls. It is the difference between an agent that suggests and one you can trust to act.
Read definitionThe EU AI Act is the European Union's risk-tiered regulation of artificial intelligence. It classifies AI systems by risk and attaches obligations — transparency, human oversight, and traceability — proportional to that risk. For work platforms deploying agents, it pushes toward designs where consequential automated actions are reviewable and attributable by default.
Read definitionHuman in the loop (HITL) is a design pattern where a person reviews, approves, or corrects an automated or AI system's actions before or while they take effect. Instead of fully autonomous operation, the system pauses at consequential points for human judgment. It balances the speed of automation with the accountability and oversight of human control.
Read definitionAn agent broker is the runtime component that sits between an AI agent and the actions it wants to take, deciding per call whether to execute inline, require human approval, or refuse. It is where propose/approve governance is actually enforced — the gate is code in the request path, not a guideline in a prompt — so an agent literally cannot run a destructive tool without passing through it.
Read definitionA tool registry is the single, authoritative catalog of the actions an AI agent may call, with each tool's name, typed arguments, and risk classification declared in one place. It is the source of truth a governance broker consults to decide whether a call is safe to run or must be gated — so capability and destructiveness are properties of the registry, not scattered per-caller decisions.
Read definitionDestructive action gating is the practice of classifying agent actions by reversibility and requiring explicit approval for the irreversible ones — delete, bulk-archive, bulk-update, external sends — while letting safe, read-mostly actions run freely. It is the operational core of agent governance: the boundary is drawn around the action, so no amount of clever prompting lets an agent quietly do something it can't undo.
Read definitionAn agent audit log is an immutable, append-only record of everything an AI agent did and proposed — each action, who or what approved it, when, and against which records — written to the same trail as human actions. It turns 'the agent did something' into an accountable, replayable history, and is what makes governed autonomy verifiable rather than merely promised.
Read definitionSkill distillation is the process of turning a finished agent session into a reusable, named skill: a model reads the transcript and extracts a structured capability — a description, a system prompt, and an allowed set of tools — so a successful one-off task becomes a repeatable behavior. It converts tribal prompt knowledge into a governed, shareable asset.
Read definitionA no-train guarantee is a vendor's contractual commitment that your data — your issues, documents, and prompts — will never be used to train its or a third party's AI models. It is the baseline trust requirement for putting proprietary work into an AI platform: your content powers your features and nothing else, with no leakage into a shared model that competitors could query.
Read definitionCapability scoping is restricting an AI agent to a defined set of tools and permissions for a given task or session, so it can only do what that role requires — and nothing more. It applies least-privilege to agents: a triage agent gets triage tools, a drafting agent gets drafting tools, and neither can reach destructive operations outside its scope, whatever a prompt might ask of it.
Read definitionDiscoverability & AI search
Generative engine optimization (GEO) is the practice of structuring content so generative AI engines — ChatGPT, Gemini, Perplexity, AI Overviews — surface and cite it in their synthesized answers. Where SEO optimizes for a ranked list of blue links, GEO optimizes to be the source a model quotes, paraphrases, and attributes when it composes a response.
Read definitionAnswer engine optimization (AEO) is optimizing content so answer engines — AI Overviews, Perplexity, ChatGPT, voice assistants — can extract a direct, accurate answer to a user's question. It focuses on being the concise, self-contained response a system lifts and reads back, rather than a page a user must open and scan to find the answer themselves.
Read definitionRetrieval & embeddings
An embedding is a numeric vector that represents a piece of content — text, image, or other data — as a point in high-dimensional space, positioned so that similar items land near each other. Produced by a model trained to capture meaning, embeddings let software compare content by semantic similarity rather than by matching exact words.
Read definitionA vector embedding is a numerical representation of text (or an image, or audio) as a list of floating-point numbers, produced by a model so that semantically similar inputs land close together in high-dimensional space. Embeddings are what let software compare meaning rather than spelling — the foundation of semantic search, retrieval-augmented generation, and recommendation.
Read definitionA vector database stores data as high-dimensional embeddings and retrieves items by similarity rather than exact match. Given a query vector, it returns the nearest stored vectors — the records most similar in meaning. Specialized indexes make this fast across millions of items, making vector databases the retrieval backbone of semantic search and AI applications.
Read definitionReranking is a second-stage refinement in retrieval that reorders an initial set of candidate results by relevance to the query, using a more precise but costlier model. A fast first pass casts a wide net; the reranker then scores each candidate against the query directly, promoting the truly best matches to the top — sharply improving the quality of retrieval.
Read definitionAn HNSW (Hierarchical Navigable Small World) index is a data structure for approximate nearest-neighbor search over vector embeddings. It organizes vectors into a layered graph that can be traversed greedily, so a query finds its closest matches in roughly logarithmic time instead of comparing against every vector — the workhorse that makes semantic search and RAG fast at scale.
Read definitionPlatform & data
An audit trail is an append-only, time-ordered record of who did what, when, and to which object across a system. Every create, edit, delete, and approval is logged immutably, so any state can be traced back to the actions that produced it. Audit trails underpin accountability, debugging, compliance, and — increasingly — oversight of what AI agents do.
Read definitionRow-level security (RLS) is a database feature that restricts which rows a query can read or modify based on the current user or context. Instead of relying solely on application code to filter data, the database itself enforces access policies on every query — a strong defense for multi-tenant systems where one workspace's data must never leak to another.
Read definitionMulti-tenancy is an architecture where one running application and database serve many independent customers (tenants), with each tenant's data strictly isolated from the others. It lets a SaaS product share infrastructure for efficiency while guaranteeing that one workspace can never see another's data — a guarantee enforced in the data layer, not left to hope.
Read definitionThe transactional outbox is a reliability pattern that guarantees a side effect — a webhook, an event, a notification — fires exactly when its database change commits, and never otherwise. The action is written to an outbox table inside the same transaction as the data change, then a separate process delivers it, eliminating the lost or phantom events that plague naive 'save then send' code.
Read definitionA webhook is an automated HTTP request a system sends to a URL you provide whenever a specified event occurs — an issue created, a status changed, a comment added. Instead of repeatedly polling an API for changes, your service receives a real-time push. Webhooks are the backbone of integrations, letting tools react to each other instantly.
Read definitionIdempotency is the property that performing an operation multiple times produces the same result as performing it once. A request that charges a card or creates a record is made idempotent by attaching a unique key, so that retries from network failures or duplicate sends don't cause double charges or duplicate data — essential for reliable distributed systems.
Read definitionOptimistic locking is a concurrency strategy that assumes conflicts are rare: instead of locking a record while it's being edited, it lets multiple users read freely and checks for conflicts only at write time, usually via a version number. If the version changed since you read it, your write is rejected so you can retry — preventing one user from silently overwriting another's changes.
Read definitionEventual consistency is a guarantee in distributed systems that, if no new updates are made, all replicas of the data will converge to the same value — eventually. It accepts that reads may briefly see stale data in exchange for higher availability and lower latency, trading the strict immediacy of strong consistency for resilience and scale.
Read definitionA single source of truth is a design principle where every piece of data has exactly one authoritative, canonical home, and all other views derive from it. Instead of the same fact living in several places that can drift apart, one record is definitive. It eliminates contradictory copies, so there is never ambiguity about which value is correct.
Read definitionTenant isolation is the guarantee that one customer's data in a shared, multi-tenant system can never be read or written by another — enforced at the data layer so a bug in application code can't leak across the boundary. It is the bedrock of SaaS trust, and in an AI platform it also means an agent's retrieval and context stay strictly within the workspace that invoked it.
Read definitionSecurity & access
Role-based access control (RBAC) governs what users can do by assigning them roles — such as admin, member, or viewer — that carry defined permissions, rather than granting rights to each person individually. It scales access management: change a role's permissions once and every user with that role updates, and onboarding becomes assigning a role, not wiring up dozens of grants.
Read definitionSingle sign-on (SSO) lets users access many applications with one set of credentials, authenticating through a central identity provider instead of a separate login per app. Sign in once and you're recognized everywhere connected. SSO improves security and user experience at once — fewer passwords to reuse or forget, and centralized control over who can access what.
Read definitionSCIM (System for Cross-domain Identity Management) is an open standard for automating the exchange of user identity data between an identity provider and the apps an organization uses. It lets IT provision, update, and deprovision accounts centrally: when someone joins, changes roles, or leaves, those changes propagate automatically to every connected application.
Read definitionOAuth is an open standard for delegated authorization: it lets a user grant one application limited access to their data in another, without sharing their password. Instead of handing over credentials, the user approves a scoped grant and the app receives a token it presents on the user's behalf. OAuth handles authorization — what an app may do — not who the user is.
Read definitionThe principle of least privilege holds that every user, service, or process should be granted only the minimum permissions needed to do its job — and nothing more. By default-denying access and granting narrowly, you shrink the attack surface: a compromised account or buggy component can only reach what it was explicitly allowed, limiting the blast radius of any failure.
Read definitionEncryption at rest protects data while it is stored — on disks, in databases, in backups — by keeping it encrypted so that anyone who gains physical or low-level access to the storage cannot read it without the keys. It complements encryption in transit, which protects data moving over the network, to give end-to-end protection across a data's lifecycle.
Read definitionSOC 2 is an auditing standard from the AICPA that assesses how a service organization handles customer data against five trust service criteria: security, availability, processing integrity, confidentiality, and privacy. A SOC 2 report, produced by an independent auditor, is the common way SaaS vendors demonstrate to customers that their controls are designed and operating effectively.
Read definitionThe General Data Protection Regulation is the European Union's comprehensive data-protection law, in force since 2018. It governs how organizations collect, process, and store the personal data of people in the EU, granting individuals rights over their data — access, correction, deletion, portability — and imposing strict obligations on data handlers, backed by fines of up to 4% of global annual revenue.
Read definitionCommonly confused
Two terms people mix up, side by side — what each means and when to use it.
From vocabulary to velocity
The words are the easy part. Planoda is where the work actually happens — free for small teams.