Why does an aging work-in-progress chart show risk that a finished-item cycle-time histogram cannot?
answer
- One chart contains only survivors
- The other looks at unfinished work
- History versus trouble today
- Stuck items never enter the histogram
- Plot days elapsed, not days taken
basics
~20 sAn aging chart plots items still unfinished against how long each has been in progress, so it surfaces trouble today. A cycle-time histogram contains only items that already finished, so the worst work — still stuck — never appears in it at all.
solid answer
~40 sA cycle-time histogram is a sample selected on completion, which makes it blind to exactly the work worth knowing about. While several items are badly stuck, the ones that do finish are disproportionately the small clean ones, so the measured average can **improve** while service gets worse; if stuck items are eventually abandoned, they never enter the data at all. An aging chart inverts the selection: every point is an item not yet finished, positioned by its current stage and by how many days it has been in progress, usually against percentile lines drawn from historical finished work. Nothing on it is an estimate — elapsed time is already known. That makes it a leading indicator you can read on any morning, rather than a report on survivors.
go deeper
Recall the difference in population: one chart is built from finished items, the other from items still in progress. Knowing that an unfinished item is absent from the histogram is the core point.
Explain the mechanics of the aging chart — current stage on one axis, days elapsed on the other, historical percentiles as reference lines — and why no estimate is involved anywhere in it.
Demonstrate the survivorship argument with a concrete case: a healthy-looking average produced partly by the absence of the stuck items. Be ready to say which chart you would open first on a Tuesday morning and why.
Own the measurement-design angle: any metric selected on completion is structurally blind to the worst outcomes. Be ready to discuss what else in an organisation's reporting has the same defect.
## The survivorship problem in flow data A cycle-time histogram is built from items that finished. That sounds obviously fine until you notice what it implies: the work that has gone worst — the item stuck for five weeks in verification, the request nobody can close because it needs a decision that keeps being deferred — contributes nothing to the chart, because it has not finished. It only enters the data on the day it stops being a problem. This produces a genuinely perverse property. During a stretch when several items are badly stuck, the items that *do* finish are disproportionately the small, clean, unblocked ones. The measured average cycle time can therefore **improve** while the service a requester receives gets worse. If some of the stuck items are eventually abandoned rather than delivered, they never enter the histogram at all, and the improvement is permanent and entirely fictional. ## What an aging chart plots instead An aging work-in-progress chart deliberately inverts the selection. Every point on it is an item that is **not yet finished**: - the **horizontal position** is the stage the item is currently sitting in; - the **vertical position** is how many days it has been in progress so far; - **percentile lines** drawn from the historical finished-item distribution — commonly the 50th, 85th and 95th — run across the chart so each unfinished item can be read against what comparable work has actually taken. Nothing on it is an estimate. The vertical axis is elapsed time, which is already known for every item without asking anyone how much is left. | | Finished-item cycle-time histogram | Aging work-in-progress chart | |---|---|---| | Population | items that completed | items still in progress | | Time frame it describes | the past | today | | Shows a stalled item | only after it finishes | immediately, and every day it stays | | Survives abandonment | no — abandoned work never appears | yes — the item sits there until removed | | Best used for | forecasting and measuring change | spotting trouble while it is still trouble | ## Reading it - **A point above the 85th percentile line** has already taken longer than most comparable work did. It is not merely late; it is outside the range the system normally produces. - **A vertical cluster in one stage** is a queue, and one that is old rather than merely large. - **An old point in an early stage** is the worst signal on the chart: something was started and then forgotten, having consumed capacity the whole time. - **The shape refreshes daily**, and points only move upward. An item that does not move right is climbing. On the construction-site safety product, the finished items from one 3-week window had a median cycle time of 5.9 days and an 85th percentile of 14.8 days — a healthy-looking histogram. The aging chart for the same morning showed two items sitting in verification at 31 and 38 days, plus one still in the first active stage at 26 days. Neither of those three appears anywhere in the histogram, and the median it reports is partly a consequence of their absence: while they sat, the only items completing were the easy ones. ## Why interviewers ask this The question separates candidates who report metrics from candidates who reason about them. The reasoning has three steps, and saying them plainly is the whole answer: 1. Any chart of completed work is a sample selected on completion. 2. The items most worth knowing about are, by definition, the ones least likely to be in that sample. 3. Therefore a leading indicator has to look at unfinished work, and the only property of unfinished work you can measure without an estimate is its age. The corollary is worth stating too: the aging chart is a **detector**, not a verdict. It tells you which items are unusual today and where they sit; it does not tell you why, and it does not tell you the two 30-day items are the same kind of problem as each other. It is also the reason a team can answer "how is it going?" honestly on a Tuesday morning, rather than only at the end of a window once the survivors have been counted.
- How can measured cycle time improve while the service a requester gets gets worse?Because the measure counts only completions. When several items are stuck, the ones finishing are the small unblocked ones, which pulls the average down. If the stuck items are eventually abandoned rather than delivered, they never join the sample, so the improvement is permanent and entirely fictional.
- What do the percentile lines drawn across an aging chart give you?A reference from the team's own history. Drawing the 50th, 85th and 95th percentiles of finished work across the chart lets each unfinished item be read against what comparable work has actually taken, so a point above the 85th line is not merely late — it is already outside the range the system normally produces.
Judging a hospital by the recovery times of discharged patients tells you nothing about the people still on the ward, and they are the ones you would want to hear about first.
saying these in an interview costs you the question
- Believes a finished-item histogram covers all work
- Reads a falling average as proof service improved
- Plots estimated effort remaining instead of elapsed age
- Ignores an old item sitting in an early stage
- Treats the aging chart as a monthly report rather than daily