Keep unfinished durations visible in a summary review

A duration summary can look shorter when unfinished items disappear from its denominator. Before calling an average representative, identify which question it answers and what remains unobserved. This fictional manual review uses invented planning turns and four identified items. The artifact is a completion-status ledger, not customer analytics or a measurement of actual work in Kiguri.

Preserve the cohort and observation cutoff

Declare one fixed cohort containing A, B, C and D. Completed durations are two turns for A, four for B and six for C. D is known to have lasted at least ten turns, but its completed duration is not supplied at the observation cutoff. These are stipulated practice records, not observations collected from a team.

Keep identity, completion status, duration or lower bound, and cutoff together. A completed duration and elapsed time for unfinished work are different kinds of evidence. Do not label D's ten as a completed duration merely to make every row numeric. Also distinguish duration from effort: the fictional record supplies no information about active work, waiting, or the reason D remains unfinished.

Label the completed-only calculation

The three supplied completed durations sum to twelve. Dividing twelve by three gives a completed-only mean of four turns. That calculation is correct for A, B and C. Its label must retain the restriction rather than becoming the average duration of all four cohort members.

Dividing the same sum by four instead gives three turns, but silently contributes zero for D. Zero is not its known lower bound and is not an observed completion. Dropping D from the record is another problem: a reader may mistake a complete three-item dataset for an incomplete four-item cohort. Keep the unfinished row visible even when presenting a valid completed-only figure.

Carry the unfinished bound into the cohort question

If all four items eventually have finite completed durations, their eventual total is at least twelve plus ten, or twenty-two turns. Their eventual mean is therefore at least 5.5 turns. This is a conservative lower bound, not an exact future mean. The supplied record gives no finite upper bound for D's completed duration or for that eventual mean.

If the cutoff establishes that D is still continuing strictly beyond ten, the bound remains conservative rather than an attained minimum. Do not claim it finishes at ten. More fundamentally, the conditional eventual mean does not promise that D will ever finish. If it never supplies a finite completed duration, the proposed finite four-item mean remains unavailable.

Retain new evidence without rewriting the cutoff

In a separately labeled what-if, suppose D's eventual duration were twenty turns. The four durations would total thirty-two, giving a mean of eight. Twenty is an invented possibility, not a completion observed at the original cutoff. Keep it apart from the current ledger and its lower-bound statement.

Have another reader explain the difference between the completed-only mean of four, the conditional lower bound of 5.5, and the what-if mean of eight. If they treat one as an estimate of the other, revise the labels. A later actual completion would create new evidence with its own observation time; it would not make that information available earlier. The finish check is an honest cohort summary with its unfinished identity and uncertainty intact, without inferring productivity, delay causes or a guaranteed completion.