Repeat Failure Rate: The Metric Nobody Tracks and Everyone Suffers

The same asset, the same failure mode, weeks apart. Repeat failures are the most expensive category of downtime because every one of them was already paid for once.

Quick answer: Repeat failure rate is the share of breakdowns where the same asset fails in the same way within a defined window, commonly 90 days. It is the cleanest measure of whether maintenance fixes causes or symptoms, and reducing it converts future reactive hours directly into capacity, because the second failure was preventable by definition.

Why repeats are the expensive failures

A first failure buys information at a price: downtime, parts, labor. A repeat failure pays the price again and throws the information away. Plants tolerate repeats for an unglamorous reason: nobody can see them. When asset records live at machine level rather than component level, the third bearing failure on the same station looks like three separate events in three months, each closed with the same temporary fix.

How to measure it

Count breakdown work orders where the same asset, at component level, failed with the same failure code within the window, divided by total breakdown work orders. Three prerequisites make the number honest: component-level asset hierarchy, failure codes that operators and technicians actually use, and automatic work-order creation so informal fixes cannot hide. Without those, the metric will report the pleasant fiction that repeats are rare.

What a high repeat rate is telling you

Three diagnoses, in rising order of difficulty. Fixes are rushed: reactive pressure means symptoms get patched at 2am and causes wait forever. Knowledge is not captured: the technician who solved it properly last year left, and the fix lives in no procedure.

Or the failure is designed in: the component is wrong for the duty, and no maintenance quality will save it, only an engineering change. The metric cannot tell you which; the video and event history behind each repeat usually can.

The repeat-failure review

Start by asking the crews. Operators can usually name the top repeat offenders before any query runs, because they clear them every week; being asked, and then seeing the repeat actually die, is the difference between a metric and a movement on the floor.

Then verify memory against data: the highest-yield 30 minutes in a maintenance month is to list every repeat from the last quarter, pick the three most expensive in lost production, and for each one answer two questions with evidence rather than memory: what did the first fix actually do, and what would a permanent fix cost against what the repeats cost?

Plants that run this review kill their worst repeats within two quarters, because once a repeat is visible and priced, leaving it alone becomes an explicit decision nobody wants to own.

Frequently asked questions

What counts as a repeat failure?

Same asset at component level, same failure mode, within your defined window; 90 days is a common convention. The window matters less than applying it consistently.

What is a good repeat failure rate?

Lower is better and zero is unrealistic. The useful comparison is your own trend and the gap between your lines; a line whose repeats keep climbing is telling you its fixes are cosmetic.

How is this different from MTBF?

MTBF averages time between all failures; it can improve while the same chronic failure keeps returning. Repeat rate isolates the preventable fraction that MTBF averages away.

Do I need software to track this?

You need component-level records, failure codes and disciplined work orders. A CMMS makes that sustainable; the metric itself is a query, not a product.

Fabrico links every stop to a component-level work order with video context, which makes repeats visible by default. Read next: reactive vs planned ratio, and PM effectiveness.