The Interruption Whitepaper

The loss with no reason

Ask a plant manager where they lose time and you'll hear about breakdowns: the failed motor, the jammed infeed, the changeover that ran long. Those are real, and they get stop reasons, reports and attention. But when we measured 62 production lines at high resolution, the largest loss on most of them wasn't the breakdowns at all. It was the relentless stutter of short interruptions (what standard OEE calls minor stops or micro-stops) between them, each one too brief to give a reason, and therefore invisible. An interruption is a gap in the count: the sensor misses at least three expected pulses, but the gap is shorter than the stop threshold. The machine may have halted, or run on with nothing to make.

We call the rhythm of those gaps the pulse of production. Across the studied footprint, the weighted-average Mean Time Between Interruptions (MTBI) was 1 minute and 14 seconds. Read that again: on average, a line ran for about seventy-four seconds before the flow of product broke. Each gap is too short to give a reason, so the time adds up unrecorded.

The fragmentation gap

It's tempting to treat a slow line as a machine problem and reach for faster equipment. The data says to look at the interruptions first. When we grouped sites by how often they were interrupted, the steadiest (MTBI around 3m 27s) averaged an OEE performance of 81.21%. The most frequently interrupted (MTBI around 1m 13s) averaged 68.67%. The 1m 14s is total run time divided by total interruptions. Busy sites contribute the most interruptions, so they pull the average down.

The gap does not mean the machines slow down between interruptions: the data does not show that. The comparison uses three-month site averages for August to October 2025: four frequently interrupted sites against three steady ones, with sites above 100% performance left out.

Three bands of operational health

Tracking the portfolio over time produced a natural, data-driven way to grade a site. Two figures set the scale, both the share of all downtime that was interruptions across all 12 sites. The 14.28% baseline is from August to October 2025, the most stable period. The 19.82% high mark is from February to April 2025, the most volatile period we measured. Each site is compared with them, and falls into one of three bands:

• Optimised: interruption ratio below 14.28%. Stable flow, high performance, clear reporting.
• Critical: between 14.28% and 19.82%. Drifting, and worth watching closely.
• Volatile: above 19.82%. Performance is at its lowest and the fragmentation gap at its widest.

The value of the bands is diagnostic. Knowing which one a line sits in tells you immediately whether it is losing capacity to fragmentation, or genuinely constrained elsewhere.

The buffer illusion

Here is where reporting quietly lies. Different lines use very different stop thresholds, the delay before a pause is logged as downtime, ranging from an aggressive 30 seconds to a lenient 5 minutes. Judge a line on its interruption percentage alone and you'll be fooled.

One line in the study reported a comfortable-looking 12.86% interruption ratio, apparently Optimised. But its actual time between interruptions was a catastrophic 15 seconds. Operators were using the generous 5-minute buffer to clear faults and manually reset the line before any stop reason was asked for. The threshold acted as an administrative mask, hiding chronic stutter and mechanical stress from every management report.

To cut through it we use a Threshold Exposure Index (TEI = MTBI ÷ stop threshold × 100). That 12.86% line scored a TEI of just 5.06%, exposing the illusion instantly. A low TEI means the reported numbers are corrupted by the buffer, and the real losses are worse than they look.

Cleaner data, sharper truth

Micro-friction doesn't just cost capacity, it corrupts analysis. Across the raw portfolio, the correlation between MTBI and reported unplanned downtime was a weak, confusing r = 0.41, because low-TEI lines had their breakdown data masked by extended thresholds. Filter to the clean, high-TEI data points and the correlation surges to r = 0.73.

That is the deeper lesson of the whitepaper: eliminating micro-friction is not only a capacity play, it is a measurement play. Remove the noise and true breakdown severity finally becomes visible, which is the precondition for fixing it. You cannot improve what your reporting is hiding.

What to do with this

The practical sequence is short. Measure interruptions at high resolution, not just breakdowns. Check each line's TEI before trusting its downtime report. Place every line in its band. Then attack the pulse, the highest-frequency stop causes, before spending a cent on faster equipment.

The named frameworks here (the pulse of production, the fragmentation gap, the three bands and the Threshold Exposure Index) are the tools we use to find where the time goes. The case studies show each one at work on a real floor.