A ten percent rework rate is not a ten percent problem

A rework rate is usually reported as a percentage, and a percentage invites the wrong arithmetic. Ten percent of the work comes back, so it costs about ten percent more. That is how it gets budgeted, and it is wrong in three separate ways at once, each of which is larger than the first.

The number to write down is not the rate. It is what the second pass consumes.

The first mistake: the second pass is not free work

An item that comes back does not cost what it cost the first time. It costs more, and the parts nobody counts are the expensive ones.

---

	The original work, done again	18 min
	Finding out it came back, and why	6 min
	Getting it back to the right person	4 min
	Checking the fix, which the first pass did not get	7 min
	One rework item, total	35 min

Illustrative figures. The original work is 18 minutes, so the second pass costs 35 minutes rather than 18: the handling around it is larger than the work itself. That ratio is typical and it is the reason a rework rate understates its own cost.

So the second pass costs about twice the first, and a rate of ten percent is already a cost of twenty percent of the original work before anything else happens. The handling is the larger half and it is the half nobody records, because no timesheet has a line called finding out.

The second mistake: it eats capacity, not just time

Twenty percent more work is only twenty percent more cost if there is somewhere to put it. If the step doing the rework is the constraint, that capacity was not spare, and every minute of it is a minute the constraint did not spend on new work.

That is not a cost increase. It is a throughput reduction, and it is priced completely differently: at what the business earns per constraint hour, which for most operations is one or two orders of magnitude above what the step costs to run. We put a number on that separately in what an hour at the bottleneck is actually worth.

If the rework happens at a step with genuine spare capacity, this second effect is zero and the arithmetic stops at the first mistake. Which of the two you are in is worth ten minutes of checking before anyone writes a business case.

The third mistake: the queue does not grow proportionally

This is the one that surprises people, and it is the reason a ten percent rework rate can feel like a crisis rather than a rounding error.

Rework adds arrivals to a step that was already busy. Utilization goes up, and waiting time does not rise in step with it: it rises much faster as the step fills, for the reasons in the piece on utilization. A step at seventy percent absorbing a rework stream can end up at eighty five, and the wait for everything going through it, including all the work that was fine, roughly triples.

That is why the complaint from the floor is never proportionate to the rate. People are not experiencing ten percent. They are experiencing the queue.

What to do with it

Count the handling, not the redo. Time one returned item end to end, including the finding out and the routing. Most operations discover the second pass costs between one and a half and three times the first, and the number is defensible because somebody timed it.

Then ask where the rework lands. If it lands on the constraint, price the recovered capacity at the constraint rate rather than the step’s operating cost, and the business case changes completely.

And fix the detection point before the defect. An item found wrong four steps later has consumed four steps of work that will be thrown away. Moving the check earlier does not reduce the rework rate at all, and it usually reduces the cost of it more than reducing the rate would.

The rate is what gets reported. What it costs is a different number, and it is always the larger one.