Original website reading companion. This original website companion is not a manuscript excerpt, verified chapter summary or account of customer results. It does not verify the book's chapter titles or numbering.
Define the question before choosing a percentage
A weekly report can show more completed work while the newest requests remain unfinished. It can also show a higher on-time percentage simply by excluding late work that has not been accepted yet. Before comparing numbers, write down the unit being counted, the population, the cutoff and the event that qualifies as completion. A label such as delivery performance is not a complete definition.
This fictional service queue covers Monday November 16 through Friday November 20, 2026, inclusive, with a review at the end of Friday. Dates represent daily records, not measured working hours. Acceptance means a recorded acceptance by the assigned reviewer; sending a draft is not acceptance. The exercise evaluates records and definitions, not employee performance. There is no measured time saving or customer outcome.
Read the ledger and preserve duplicate evidence
A and B each arrived in two source records but refer to one work item apiece. C, D, E and F each arrived once. That makes eight incoming source records and six distinct new work items. Preserve both records for A and B as history, but link each pair to its work identifier before counting workload. Do not delete a message merely to make a dashboard total agree.
G and H were already open before Monday. Both were accepted during this week. They belong in period throughput and the backlog reconciliation, but not in the cohort of work first received this week. F was cancelled on Friday before acceptance. Keep that cancellation visible as a separate event; treating it as completed would answer a different question.
B needed one revision before its Wednesday acceptance. That revision is a quality event attached to B, not a second completed work item. No revision is recorded for the other accepted items. A blank or zero event count in this fictional ledger does not prove that a real process has no defects; a real report needs a defined method for recording returns.
| Item | Received | Due | Accepted | Cancelled | Source records | Revisions |
|---|---|---|---|---|---|---|
| A | 2026-11-16 | 2026-11-18 | 2026-11-17 | — | 2 | 0 |
| B | 2026-11-16 | 2026-11-18 | 2026-11-18 | — | 2 | 1 |
| C | 2026-11-17 | 2026-11-19 | 2026-11-20 | — | 1 | 0 |
| D | 2026-11-18 | 2026-11-20 | — | — | 1 | 0 |
| E | 2026-11-19 | 2026-11-24 | — | — | 1 | 0 |
| F | 2026-11-20 | 2026-11-24 | — | 2026-11-20 | 1 | 0 |
| G | 2026-11-13 | 2026-11-17 | 2026-11-16 | — | 1 | 0 |
| H | 2026-11-13 | 2026-11-20 | 2026-11-19 | — | 1 | 0 |
Scroll the ledger horizontally on a narrow screen. Download the fictional work ledger CSV.
Separate this week's arrivals from this week's completions
The six distinct arrivals are A through F. Excluding the explicitly cancelled F leaves five non-cancelled arrivals: A through E. Three of those five were accepted by Friday: A, B and C. The cohort completion fraction is therefore 3/5, or 60%. D and E remain open. Report that denominator and the separate cancellation count alongside the fraction.
Period throughput is five accepted items: A, B, C, G and H. It includes two opening-backlog items and three new arrivals. Dividing five period completions by five non-cancelled arrivals would produce 100%, but it would mix populations and would not show that every new arrival was completed. Keep throughput as a count unless a meaningful comparison population has been specified.
Keep unaccepted due work in the on-time denominator
Four non-cancelled new items had deadlines on or before Friday: A, B, C and D. A and B were accepted by their recorded deadlines. C was accepted Friday against a Thursday deadline. D was due Friday and remains unaccepted at the cutoff. The due-cohort on-time fraction is 2/4, or 50%. E is due the following Tuesday and is outside this denominator; it is not late at Friday's cutoff.
Among the three completed new items, two were on time: 2/3, approximately 66.7%. That statistic answers how many accepted arrivals were on time. It leaves out due-but-unaccepted D and must not be labeled as the due-cohort on-time result. Neither fraction establishes a trend from a single fictional week, and neither explains why the late or open item occurred.
Reconcile the backlog before interpreting the dashboard
Opening backlog is two items, G and H. Add six distinct arrivals, subtract five acceptances and subtract one cancellation: 2 + 6 - 5 - 1 = 2 open items at the cutoff. Those two items are D and E. This identity depends on a consistent work-item boundary and on acceptance and cancellation being mutually exclusive closing events in this exercise. Reopened or split work would require additional, explicitly defined events.
Only D is overdue at the Friday cutoff. E is open but not yet due. Calling both items late would misstate the ledger. The data also cannot support an average cycle time for all five non-cancelled arrivals: D and E have no acceptance date. You may describe the completed subset with that limitation, but do not replace missing acceptance dates with Friday and present the result as measured completion time.
Use a review question, then test the recording rule
A useful next review asks why D lacks acceptance and whether its recorded deadline and scope are still the approved ones. Assign the next authorized reviewer, retain the source references and record any approved change rather than editing history to improve the percentage. For B, inspect the revision reason before deciding whether a procedure needs clarification. A count identifies a place to investigate; it does not supply the cause.
Before adopting a real weekly report, test a duplicate arrival, a cancellation, an old item accepted this week, an overdue unfinished item and an open item due later. Confirm that each enters the intended population once. Keep the denominator definition with the report so another reviewer can reproduce the result. The downloadable ledger contains only fictional records and can be used to check those definitions without sharing personnel or customer information.
Review a blocked decision queue or document a recurring workflow. Return to the Built to Run overview for purchase options and other website companions.
Original website exercise by Robertson Books. Reviewed 2026-11-22.