Glossary
Metriq describes how work flows using a small, precise vocabulary borrowed from flow science (and a few terms of its own for the agent loop). If a word on a chart or forecast is unfamiliar, look it up here. Every term has a stable anchor, so other pages can link straight to the definition.
A note before you start: Metriq is estimate-free by design. You will not find story points, t-shirt sizes, or guessed durations anywhere in this list. Forecasts come from measured flow — how long real work has actually taken — run through a Monte Carlo simulation. The terms below are what make that possible.
Cycle time
The time a single piece of work spends actively being worked — from the moment it starts (enters an in-progress column) to the moment it is done.
Cycle time is the engine behind every forecast. Because Metriq records when each item starts and finishes, it never has to ask you to guess how long the next item will take — it looks at how long your last items actually took. Shorter, steadier cycle times make forecasts tighter and more trustworthy.
Lead time
The total time a piece of work spends in the system — from the moment it is created (enters the backlog) to the moment it is done.
Lead time is what a stakeholder experiences: "I asked for this; how long until I get it?" It includes waiting in the backlog, so it is always equal to or longer than cycle time. Use lead time to set expectations with people outside the team; use cycle time to understand the team's working pace.
WIP (Work In Progress)
The number of items being worked at once — the count of cards sitting in the in-progress columns of the board right now.
Metriq supports WIP limits on purpose. Less work happening in parallel means each item finishes faster, cycle times stay short, and forecasts stay reliable. A limit comes in two modes: warn (the default — you get a "WIP limit exceeded" warning but the move still goes through) and block (the move is refused server-side until something moves out). Either way the friction is the point. Watch for aging WIP: items that have been in progress longer than usual, a leading signal that flow is stalling.
SLE (Service Level Expectation)
A commitment about how reliably work finishes within a target time — stated as a percentile plus a number of days. For example, "85% of our items finish within 7 days."
An SLE is not a deadline and not an estimate. It is a probabilistic agreement read directly from your historical cycle times: if 85% of past items finished inside 7 days, then "85% within 7 days" is a promise you can actually keep. Metriq proposes a starting SLE (commonly the p85 of your data) and shows you, item by item, whether you are tracking to it.
p85
The 85th-percentile value of your cycle times: the duration that 85% of your completed items finished within (and 15% took longer).
Percentiles are how you read a forecast honestly. Saying "items usually take about 4 days" hides the risk in the slow 15%. Saying "85% of items finish within 7 days" (p85) tells you the level you can plan around with confidence. Metriq plots several percentile lines — p50 (the median, a coin-flip), p70, p85 (a common planning level), and p95 (near-worst-case). A higher percentile is a safer, more conservative promise. p85 is the usual default because it balances ambition against reliability — exactly the level most teams adopt for an SLE.
Monte Carlo
The simulation technique Metriq uses to turn your team's throughput — how much work it actually completes — into a forecast, by replaying thousands of possible futures and counting how they land.
Rather than guess a single delivery date, Metriq samples from your team's actual past throughput over and over, simulating thousands of ways the remaining work could play out (each run replays your real throughput until the work is finished, then sees how long that took). The spread of those simulated finishes becomes the forecast: instead of one fragile date, you get a distribution you can read at a chosen percentile. This is why a forecast is stated as "p85 = next Thursday" rather than a promise — it is a calibrated bet built from evidence, not an estimate. (It needs a baseline of completed history before it can run.)
Dispatch
The act of handing a work item to an AI agent to work on — the start of the agent loop.
When you dispatch an item, Metriq records the moment, opens an agent session, and tracks the work as it progresses. A dispatch is a request for an outcome, not a guarantee of one: the loop only closes when the agent reports back an artifact and you bless it.
Agent / session
An agent is an AI worker (for example, a Claude-powered coding agent) that can pick up dispatched items and work them autonomously. A session is one continuous span of an agent working — opened at dispatch, closed when the agent finishes or stops.
Sessions are how Metriq keeps the agent loop accountable: each one is timed, attributed to a specific item, and produces a trail you can inspect. Agent work flows through the same board, WIP limits, and cycle-time accounting as human work — there is no separate, unmeasured track.
Bless
Your explicit approval that an agent's delivered work is good — the act that closes the agent loop.
The loop is dispatch → artifact → bless: you dispatch an item, the agent does the work and reports back an artifact (a link to what it produced — a pull request, a document, a deploy), and then a human blesses it. Blessing is deliberate: an item cannot be blessed without an artifact to point at, so approval always references real output rather than a status anyone simply claimed. Bless is the human checkpoint that keeps autonomous work honest.
Board
Metriq's single work surface: four fixed columns — Backlog, Ready, In Progress, and Done — that every item moves across, left to right.
The board is deliberately not customizable. The four columns give every team the same, consistent definition of when work started and finished, which is exactly what makes cycle time comparable from one card to the next and keeps the Monte Carlo forecast trustworthy. For the reasoning behind the fixed shape, see Why four fixed columns.