How to measure rework without counting the same item twice

01

Define what counts as rework

Choose a process and a review period. Define rework as a return to an earlier step to correct work that did not meet an agreed requirement. Distinguish it from a customer-requested change, a planned review and a cancelled item. Record uncertain cases for review.

For each item, keep its ID, work type, return date, reason and originating step if known. One item may have several return events; preserve each event while linking all of them to the same item.

02

Choose a denominator and outcome window

For an item-based rate, divide the number of distinct completed items requiring at least one rework event by the total distinct completed items in the same cohort. State when the observation window closes so late corrections are handled consistently. Count return events separately when you want to understand repeat corrections.

Hypothetical example: of 80 items completed in a month, 12 require rework within a defined 30-day window. The item-based rate is 12 divided by 80, or 15%. Those 12 items generate 18 return events, so the event count is 18. It is not a 22.5% item rate, because some items returned more than once.

03

Investigate causes and effort

Group returns by reason and work type, then review examples with the people who perform and receive the work. Estimate correction time separately from original processing time and note missing records. A high count alone does not establish the cause or the cost.

Try a specific prevention step, such as clarifying a requirement at intake, on a defined group. Compare the same rate and observation window before and after, while tracking completion time and quality. Look for new problems created by the change.

Suppose six of the 12 returned items in the example lacked the same required detail. Review their records to see where that detail should have been supplied and whether the requirement was clear at the time. If the team adds an intake check, record how often it finds a missing detail and whether customers wait longer while it is collected. A correction may be prevented earlier without eliminating the effort needed to obtain accurate information.

Allow the full 30-day window to close before comparing cohorts. Items completed near the end of the month need the same opportunity for a late return as items completed at the start. If returns are discovered after the window, report them separately and consider whether a longer window better reflects the service.

04

Keep item and event counts separate

Give each correction an event ID and attach it to the original item ID. Then report two figures: distinct items with at least one correction divided by eligible completed items, and total correction events. The number of correction events can exceed the number of affected items when an item is corrected more than once. Keep estimated correction hours in a separate field so the count and labor estimate remain distinguishable.