The same headline equipment score can hide different losses.
All notes

Operational excellence

Same OEE, opposite fixes: turning the percentage back into minutes

Two machines can report the same OEE and need completely different work. Here's how I turn the score back into minutes a team can act on.

On the Unilever lines I worked on, OEE moved from roughly 60% to roughly 75%. To choose the next improvement, we had to trace the lost minutes behind the percentage.

The shift example below shows how the same OEE can hide different losses and lead to different work.

A quick refresher on OEE

Overall equipment effectiveness multiplies three factors (Vorne; the same structure appears in the ISO 22400 KPI standard, summarised by NIST):

Factor Question it answers Formula
Availability Was the machine running when it was supposed to? Run time ÷ planned production time
Performance When it ran, did it run at its ideal speed? (Ideal cycle time × total count) ÷ run time
Quality Were the parts right first time? Good count ÷ total count
OEE How much of the planned time made good parts at full speed? Availability × Performance × Quality

To choose an action, separate the three factors and inspect the losses behind each one.

One shift, worked through

Here's an illustrative packaging shift, the same example we use in the free OEE calculator on the Vrolen site:

Input Value
Planned production time 450 min
Stopped 60 min
Ideal cycle time 0.5 min per unit
Units produced 720
Good units, first pass 690

That gives 86.7% availability, 92.3% performance and 95.8% quality, so 76.7% OEE. Convert that result back into the planned minutes:

450 planned minutes, one illustrative shift 345 min fully productive 60 min stops (Availability) 30 min slow running, small stops (Performance) 15 min rejects and rework (Quality) OEE = 345 / 450 = 76.7%
Where the shift went. The score is 76.7%; the minutes tell you what kind of problem you have.

Sixty minutes of stops, thirty minutes of slow running and small stops, fifteen minutes of rejects. That loss profile starts a far better conversation than "we need to raise OEE by five points".

Same score, different problem

Compare two illustrative lines with the same 75.2% OEE:

Machine Availability Performance Quality OEE
A 80% 95% 99% 75.2%
B 95% 80% 99% 75.2%
  • Machine A spends too much planned time stopped. The work is on breakdowns, changeovers and waiting for material.
  • Machine B runs almost all shift but loses rate: slow cycles, microstops, starving or blocking from its neighbours, maybe a cycle time nobody has validated in years.

If both teams get the same instruction to "improve OEE", neither of them has actually been told anything. Researchers have made the same point for years: OEE is useful for spotting losses, but equipment-level OEE on its own isn't enough when machines work together as a line (Muchiri & Pintelon, 2008; Braglia, Frosolini & Zammori, 2009).

What moved the number for us

At Unilever the work that moved OEE was mostly unglamorous, and it lined up with the loss types above:

Loss What we worked on What it did
Stops (availability) Changeover reduction (SMED), autonomous maintenance Changeovers roughly halved, from about 90 to about 45 minutes
Speed (performance) Standard work, line-side material flow Fewer small stops and waits for material
Quality In-process checks, first-pass yield work Defects caught on the line, not at the end of it
Visibility A downtime and loss dashboard 60% less time spent preparing production reports

We tackled those losses through ten-plus smaller projects within the World Class Manufacturing program. The dashboard gave each shift the same breakdown of lost time.

Four questions before you pick the next project

Even the biggest loss isn't automatically the best thing to fix. Before committing a team, I ask:

  1. Which factor actually moved? Availability, performance or quality.
  2. Which events created the loss? Not the category, the actual stops, jams and rejects.
  3. Is this machine the constraint right now? Recovering minutes on a machine that isn't holding the line back mostly makes a bigger queue.
  4. Which result would the change improve? Check good output, lead time and delivery, alongside any safety or quality requirement that needs attention.

That third question is where OEE and flow meet, and it's the subject of my note on why busy machines make customers wait.

Where software can help

This is the logic I'm building into Vrolen (more about it here): connect the score to the events underneath it, check whether the loss sits at the constraint, and test a countermeasure on a model of the line before anyone changes the real one.

Vrolen's guided what-if comparing cautious, balanced and ambitious options
Testing a countermeasure in Vrolen before touching the line. Cautious, balanced and ambitious options, side by side.

For the next shift review, take the stop records, speed losses and reject quantities behind the score. Check the minutes, then investigate the loss at the current constraint. Vrolen’s guide to why higher OEE can leave output flat can help narrow down what to check.

OEE Explained: Availability, Performance and Quality (Worked Example) A separate bottle-filling run: 80% availability, 90% performance and 95% quality give 68.4% OEE.

Sources

The shift and machine figures are illustrative. The Unilever figures are approximate and from my own work.

NextDigital transformation strategy: start with the work

Discuss an operations problem

I work with operations and transformation teams on operational excellence, digital transformation programs, supply chains and industrial AI. Send me the problem you’re working on and the result you need.

Abolfazl ShirkavandImage preview unavailable