Module 1 of 2 · 120 minutes
Availability, performance and quality: measuring a line honestly
By the end of this module you will be able to
- Calculate overall equipment effectiveness from shift data
- State what each of the three factors measures and how it is inflated
- Decide what belongs in loading time and defend that decision
- Map the six big losses onto the factor each one damages
- Use mean time between failures and mean time to repair to choose a fix
Amara Every plant quotes an OEE. Why do you distrust the number?
Nadia Because the formula is trivial and the definitions are not. Three fractions multiplied together. Anyone can do the arithmetic. What decides the answer is what you put in loading time, and that is a judgement somebody made once and rarely wrote down.
Amara Start with the arithmetic anyway.
Nadia Availability, performance, quality, multiplied. Ninety per cent on each looks respectable and gives you seventy three. That multiplication is why plants that feel busy find a quarter of their intended production time produced nothing they could sell.
Amara Loading time. What is the argument?
Nadia What did you intend to run. Nobody counts Christmas Day. After that it is contested. Planned maintenance, changeovers, trials, a shift with no orders. Argue any of those out of loading time and the number rises without one thing changing on the floor.
Amara So which convention is right?
Nadia Either. Neither is wrong. Excluding planned maintenance answers how well the equipment ran while you meant it to run. Including it answers how much of the calendar became product. Pick one, write it down, and never move it because a month looks poor.
Amara How often does that happen?
Nadia Often enough that it is the first thing to check when OEE improves sharply with no corresponding change in output. If the tonnage did not move and the number did, somebody edited a definition.
Amara Performance. You called two of its losses invisible.
Nadia Reduced speed and small stops. Reduced speed is a line running below its demonstrated best because it jams at full rate, and that decision is usually so old that nobody remembers taking it. Small stops are the ten second clearances.
Amara Ten seconds. Does that matter?
Nadia Two hundred times a shift it matters enormously. It is frequently the largest single loss in the plant, and it is almost never in the downtime log, because filling in the log takes longer than the stop took. That is not laziness. It is a measurement system asking for something impossible.
Amara How do you catch them, then?
Nadia You stop relying on people to type them. Count from the line itself. Any gap between units beyond the cycle time is a stop, whether or not anyone declared it. The machine is a more reliable witness than a clipboard at the end of a twelve hour shift.
Amara Ideal cycle time. People set that generously.
Nadia Constantly, and it destroys the factor. If ideal cycle time is the rate you usually achieve, performance is ninety eight per cent by construction and tells you nothing. It has to be the demonstrated best the equipment has ever sustained.
Amara Quality. Where does that one go wrong?
Nadia Rework. Quality means right first time. A unit that went to a rework bench and came back consumed capacity twice, and the whole point of the factor is to make that visible.
Amara But it sold in the end.
Nadia It did, and the plant paid twice for it. Where rework rejoins the good count you get quality above ninety nine per cent in a plant that is running a permanent rework cell. The number is denying the existence of a room you can walk into.
Amara The six big losses. Are they more than a poster?
Nadia They are useful for one specific reason. Each loss lands in exactly one factor. So an improvement makes a falsifiable claim: fix changeover and availability must move. If you cut changeover time in half and availability is unchanged, the saving went somewhere else and you should find out where before you celebrate.
Amara Give me an example of that going wrong.
Nadia A team halves changeover and availability does not move, because the line now sits idle waiting for materials that were never the constraint before. The improvement was real. The benefit was absorbed by the next problem, which is normal and worth knowing.
Amara Mean time between failures and mean time to repair. Why both?
Nadia Because they are different diseases with the same symptom. Both dent availability. A short time between failures with a fast repair means the machine fails constantly and your team has become superb at fixing it. That is not a compliment, it is a warning.
Amara And the reverse?
Nadia Long time between failures, slow repair. Rare events, no spare on the shelf, nobody on shift who has done the job before. Completely different project. One is engineering out a fault, the other is stores and training.
Amara If somebody has one hour a week, where do they spend it?
Nadia On the downtime Pareto, and on making the counting automatic. Most plants improve fastest not by fixing anything but by finding out what is actually stopping them, because the thing everyone assumes is the problem is usually third on the list.
The written material
What the measure is for
Overall equipment effectiveness answers one question: of the time you intended to produce, how much produced good output at the rate the equipment is capable of. It is three fractions multiplied together, and the multiplication is the point. Three respectable looking factors of ninety per cent each give an OEE of seventy three per cent, which is why plants that feel busy discover that roughly a quarter of their intended production time produced nothing saleable.
Around eighty five per cent is generally treated as world class for a discrete line. A typical plant that has never measured seriously lands somewhere in the fifties or sixties. The gap is not usually one dramatic failure. It is a hundred small stops nobody logged.
Availability, and the loading time argument
Availability is run time divided by loading time. Run time is when the line was actually producing. Loading time is the time you intended it to produce, which is where every serious disagreement about OEE begins.
Nobody counts a shift the plant was closed. Beyond that it gets contested quickly. Planned maintenance, changeovers, trials, training, a shift with no orders: each of these can be argued out of loading time, and each removal raises the number without anything changing on the floor.
The defensible position is to state the convention, write it down, and never change it to make a month look better. A plant that excludes planned maintenance is answering how well the equipment ran while it was meant to be running. A plant that includes it is answering how much of the calendar turned into product. Both are real questions. Only one figure at a time can be on the board.
- Breakdowns and unplanned stoppages: always a loss
- Changeover and setup: a loss, and the one most worth attacking
- Planned maintenance: excluded or included by stated convention
- No orders, no materials, no operator: excluded only if you never intend to hold the line accountable for it
- Trials and first article runs: usually excluded, always declared
Performance, and the losses nobody logs
Performance compares what the line produced against what it should have produced in the time it was running. Total count multiplied by the ideal cycle time, divided by run time.
Ideal cycle time has to be the equipment's demonstrated best, not the rate the line usually manages. Setting it to the usual rate guarantees a performance figure near one hundred per cent and removes the entire diagnostic value of the factor.
Two losses live here and both are close to invisible. Reduced speed is a line run deliberately slower because it jams at full rate, which is a decision so old that nobody remembers making it. Small stops are the ten and twenty second interruptions to clear a misfeed. Individually beneath notice, collectively often the largest single loss in the plant, and almost never in the downtime log because logging one takes longer than the stop did.
Quality means first time
The quality factor is good count divided by total count, and good means right the first time. A unit that was reworked is not good output for this purpose, however saleable it eventually became.
This is the factor most often quietly corrupted, because rework usually happens away from the line and the reworked unit rejoins the count later as though nothing happened. The capacity it consumed twice disappears from view, and so does the reason it failed. Plants that count rework as good output routinely report quality factors above ninety nine per cent while running a permanent rework cell whose existence the number denies.
First pass yield is the honest name for the same idea, and it travels better across a business because finance recognises it too.
The six big losses
The losses are a checklist, not a theory. Each one lands in exactly one factor, which tells you which number should move if you fix it. An improvement that does not move the factor it claims to have improved has not been proven.
Reliability and recoverability are different problems
Mean time between failures is total run time divided by the number of failures. It measures how often the equipment lets you down. Mean time to repair is total repair time divided by the number of repairs. It measures how long you take to get going again.
They point at completely different work. A short mean time between failures with a fast repair is a reliability problem: the machine is failing constantly and the team has become excellent at fixing it, which is not a compliment. A long mean time between failures with a slow repair is a preparedness problem: rare failures, but no spare on the shelf and nobody on shift who has done it before.
Both feed availability, so both show up as one dent in OEE. Reading only the OEE hides which of the two you have, and the two have almost nothing in common as improvement projects.
Knowledge check
The knowledge check and your certificate need a free account, so that your progress and results can be saved as evidence.
The learning itself stays free and open. You are reading all of it right now without an account.