Ask what “a request takes ten days” actually means

A project can begin with a statistic that sounds decisive: the average request takes ten days, so the organisation wants software that brings it down to three. Before estimating the build, inspect what the statistic describes. A summary can be arithmetically correct while answering a different question from the one the team thinks it answers.

This fictional example concerns administrative specification requests. The durations are invented and do not describe Faith Forge Labs performance, a customer result or a promised response time.

Establish the records included

The operations lead supplies all six requests completed in a defined demonstration period. The duration starts at the agreed acceptance event and ends when the requested specification is released. Those boundaries are the same for every record. Open requests are outside this particular completed-request measure and remain listed separately.

The six durations, sorted from shortest to longest, are 2, 2, 3, 3, 4 and 46 elapsed days. Their sum is 60 days. Dividing by six gives a mean of ten days. There is no missing record or arithmetic fault in that calculation.

The team initially describes the ten-day figure as the experience of a typical request. Looking at the members reveals that five of the six completed within four days. The sixth requires a different discussion.

Keep two different questions visible

For this six-member set, the median is the average of the third and fourth ordered values: three days. It describes the middle of the observed completed-request distribution. It does not establish that every new request will finish in three days, nor does it explain why the longest one took 46.

The mean remains useful when the question concerns the total elapsed duration distributed across these six records. It is not a measure of staff effort: a request can wait for information without someone actively working on it throughout the interval.

The first conversation therefore distinguishes three questions. What is the middle duration in this defined set? Which records have longer elapsed durations? What work or waiting caused those differences? A single number cannot answer all three.

Inspect the long case without deleting it

The team opens the 46-day record, R-6. Its history shows that a required specification from a supplier was unavailable for part of the interval. The request remained within the approved measurement boundary, so the analyst does not remove it simply to make the result look better.

The evidence supports a narrower planning question: can staff find requests awaiting supplier information, identify the owner and see the next review action? It does not establish that automation would have made the supplier deliver the missing specification sooner.

The team also reads one two-day and one four-day record. Their ordinary completion steps do not reveal the same waiting condition. This prevents an exceptional case from becoming an assumed cause of delay across the whole set.

Choose a first result worth building

The organisation rejects a proposal framed only as “reduce average time to three days”. That outcome is neither supported by the six examples nor wholly controlled by the application.

Instead, it approves a small reporting and review improvement. The view will label the mean and median separately, show the six-record population and retain a route to individual long cases. The operations lead will use a separate queue for requests waiting on external information, with an owner and a next review decision.

The proposal does not discard the existing request system. Its first scope is to make the evidence behind the summary usable and expose the relevant follow-up task. Any later automation must be justified by what staff observe in that task.

Record what the demonstration resolves

In the fictional acceptance review, the analyst reads back six completed records, a mean of ten days and a median of three. R-6 remains present at 46 days. Opening it shows the supplier-information wait and the assigned follow-up owner. The two-day control is not placed in that waiting category.

A separate empty-period test displays no calculated duration, rather than zero days. An even-sized four-record control of 1, 2, 8 and 9 produces a median of five, proving that the display does not simply select the lower middle value. The labels state the period, the completed-request population and the elapsed-day unit.

These checks accept the reporting interpretation and demonstrated review route. They do not claim a future reduction or a representative statistical study of the organisation's entire operation.

Bring the records behind an important summary to the first project discussion. The measurement service can help connect a metric to a decision, while the project workbook keeps the receiving task in view.