Cycle Time

How long finished work actually took, door to door — the full spread of delivery time over a rolling window of sprints, not just an average, so the slow outliers that actually hurt stay visible.

What you see

• Filter chips for three kinds of work — Planned Work, Bugs, Exploration — multi-select, with at least one always on.
• A window selector: the last 3, 6 (default), or 12 sprints.
• A confidence note next to the filters, based on how many items are in the current sample.
• A four-tile reading strip: Typical (p50), 85th percentile, Items delivered, Sprints covered — each compared against the previous window of the same length.
• A scatterplot, "Cycle time by sprint," with an "Export CSV ›" action — one dot per delivered item, when it finished and how many days it took.
• A "Delivered items" table beside it, sortable by completion date or by days.

How we calculate it

End (completion)

An item counts as complete only once its current state is mapped to Done under your workflow mapping. An item that isn't currently in a mapped-Done state simply doesn't appear on this chart — it isn't shown as "in progress" here, it's just not part of a chart about finished work.

Start

The first time the item ever moved into a state mapped as in progress. If that never happened, its Activated Date is used instead; failing that, its creation date is used and flagged as an estimate, since a creation date measures the item's age, not real working time — the true cycle time is likely shorter than shown.

Reopened or backward-moving items

The start stays anchored to that item's very first activation, even if it was later reopened or moved backward through the workflow — a full rework cycle is included in the measured span, not reset by the reopen.

Cycle time value

Fractional calendar days from start to completion — weekends included, not configurable to working days anywhere in this view. Items spanning more than about two years, or with a negative span, are dropped from the chart entirely rather than shown as a misleading outlier.

Percentiles (P50 / P85)

Nearest-rank, not interpolated — the value shown is always a real cycle time some actual item had, never a number in between two items. Percentile reference lines on the chart only appear once there are at least 10 delivered items in the current window; below that, a note explains why no lines are drawn.

Confidence note

Fewer than 15 items is a low-confidence sample, 15–29 is medium, 30 or more is high — the same bands used everywhere else in Agile Gauge.

Duplicate items

An item appearing in more than one sprint in the window is only counted once.

Scatterplot

The x-axis is the item's real completion date; the y-axis is its cycle-time days. Up to three colours distinguish Planned Work, Bugs and Exploration — capped at three because a fourth colour on a scatterplot becomes unreadable. Clicking a percentile line lists every item at or above that line.

What you can click through to

• Every dot and every table row opens that item's detail, its journey through every state it actually passed through, and a link to open it directly in Azure DevOps.
• The P50 and P85 reference lines are themselves clickable, opening the list of items at or above that day count.
• "Export CSV" downloads exactly the currently filtered and sorted table rows.