Rework rate is the share of units that fail the first time and have to be reworked, repaired, or reprocessed before they pass. Rework rate = units reworked ÷ units produced. It is the metric that surfaces a loss OEE and final yield both hide, because a reworked unit eventually passes, so it quietly counts as good.
That is the whole problem with rework. It leaves no scrap in the bin and no gap in the shipment, so it disappears from the numbers that are supposed to catch it. A line can report excellent yield and a healthy OEE while a rework crew works full-time behind it, fixing a third of everything the process makes wrong the first time. A separate rework-rate metric drags that hidden work into daylight. This guide covers how rework hides, how to calculate the rate, what it really costs, and how to bring it down.
What is rework rate and how is it calculated?
Rework rate is the percentage of produced units that required rework, any repair, reprocessing, adjustment, or retest after a failed check, before they could pass. The formula is simple: rework rate = units reworked ÷ total units produced, expressed as a percentage. Some plants track it in units, others in labor hours spent on rework; both are useful, and they answer slightly different questions. Units tell you how often the process misses; hours tell you how expensive each miss is to recover, which matters when one defect takes a minute to fix and another takes an hour.
A worked example with hypothetical numbers. A line produces 5,000 units in a week. 150 are scrapped outright, and 400 fail a check, go to the rework bench, and pass on the second attempt. The rework rate is 400 ÷ 5,000 = 8%. Note that the 150 scrapped units are not rework, they were never recovered, so scrap and rework are separate categories that should be counted separately, even though both are quality losses.
Why does rework hide inside OEE and yield?
Because both metrics forgive it. Final yield counts every good unit that ships, including the ones saved by rework, so a reworked unit lands in the good column and the yield number never flinches. OEE is stricter, its Quality factor counts only first-time-good units, so rework does lower it, but rework shows up there as a small blended dent inside one percentage, not as a number anyone can act on. Neither metric tells you how much rework there is, what it costs, or where it comes from.
This is why first pass yield and rework rate are complementary. FPY tells you the process is missing on the first attempt; rework rate tells you how much of that miss is being recovered by hand rather than scrapped. Together they size the hidden factory, the inspect, rework, retest loop that consumes labor, floor space, and schedule while producing nothing new. Rework is the part of that loop that is easiest to ignore, precisely because it ends in a good unit.
There is a second reason rework stays invisible: it often lives outside the standard flow entirely. A defect gets pulled aside to a bench in the corner, fixed by an experienced hand who does not stop to log it, and slipped back into the next pallet. No transaction, no record, no number. The unit ships and the paperwork shows a clean run. Multiply that quiet recovery across a shift and a plant can be reworking hundreds of units a week while every report it produces says the process is running fine. The rework rate exists to close that gap, but only if the rework is actually captured, which is where most of the effort really goes.
What does rework actually cost?
Far more than the visible labor at the bench. A reworked unit is paid for at least twice: once to make it wrong, and again to make it right. On top of the direct rework labor sit the costs of handling and moving the unit back, re-inspecting it, the capacity the rework station steals from other work, and the schedule slip while the unit sits in the loop. None of that shows up as scrap, so none of it is counted unless you count it on purpose.
The American Society for Quality reports that the total cost of poor quality, appraisal, internal failure like scrap and rework, and external failure, commonly runs 15–20% of sales revenue, reaching as high as 40% at poor performers. Rework is a core piece of the internal-failure share, and it is the piece a plant is most likely to treat as a normal cost of doing business. A rework rate makes it a line item instead of a habit, which is the first step in the broader cost of quality conversation.
What is a good rework rate?
Lower than you think, and measured against your own trend rather than a headline number. The first honest rework rate a plant measures is almost always higher than anyone guessed, because most rework was never logged. Rather than chase a benchmark, watch the direction: a rework rate that is falling week over week, with the causes being removed one at a time, is worth more than a low number nobody trusts.
Two cautions. First, do not confuse a low rework rate with a low scrap rate, a plant can cut rework simply by scrapping the same defects instead, which is worse, not better. Track scrap and rework together so you see the whole quality loss and not a shell game between two bins. Second, rework rate is not a people metric. If it is used to grade operators, the rework will keep happening and simply stop being logged, and you will have traded a visible problem for an invisible one. Rework rate measures the process; keep it pointed there. It belongs on the board alongside the rest of your manufacturing KPIs not in a performance review.
How do you reduce the rework rate?
Reducing rework means finding the defects that cause it, removing the biggest causes, and refusing to let recovery work hide the miss. A dependable sequence:
- Log every rework event where it happens. A rework rate is only as honest as the logging, and a ten-second touch-up at the station takes longer to write down than to do. Capture it at the point of work or it never gets counted.
- Attach a defect code to every reworked unit. Rework without reasons is a cost with no cause. A short, shared code list feeding defect tracking turns the rework rate into a ranked list of problems to solve.
- Pareto the codes and attack the top one. Most rework traces to a few defects. Work the largest with root-cause analysis and statistical process control on the variable that drives it.
- Mistake-proof the top failure mode. Where the defect is a settable parameter or a human error, design it out with a poka-yoke or an interlock so the unit cannot be made wrong in the first place.
- Move detection upstream. A defect caught at its source is cheaper to fix than the same defect caught three stations later with value already added. Put the check where the failure happens.
- Re-measure and hold the gain. Post the rework rate, watch the worked code shrink, then move to the next. Improvement that is not re-measured drifts back within a month.
How does rework rate connect to the rest of the plant?
Rework is where a quality problem becomes a capacity problem. A reworked unit consumes the line twice, so a high rework rate quietly drags on throughput and shows up in the OEE Performance and Quality factors at once, the rework station steals capacity and the failed unit dents quality. It also feeds the time ledger: defect investigations and rework holds land in the machine downtime log as quality stops. A rework rate that is trending down usually pulls throughput and OEE up with it.
The one thing rework rate depends on completely is honest capture, and that is exactly where paper fails: the fastest way to make a rework rate look good is to not write the rework down. Plants that capture rework digitally at the station, the way Harmony turns paper checks and quality logs into live, searchable records feeding real-time boards and production reporting (see the platform), find their true rework rate within weeks. It is usually higher than the reported one, which is the point: you cannot reduce a loss you have arranged not to see. The CLS case study shows what moving from paper logs to real-time capture looks like on the floor.