Use case · OEE & throughput

OEE calculated from machine data. Not from the end-of-shift sheet.

Availability, performance and quality come from tags your machines already expose. Meddle turns them into an OEE that is comparable across lines and shifts, and surfaces the micro-stops nobody records because they last less than a minute.

Line 2 · morning shiftOEE and components
Availability91%
Performance78%
Quality97%
Micro-stops (< 60s)142 events
Shift OEE68.8%
Why is line 2 performance below 80%?
  • Automatic
    OEE computed from tags, with no manual entry
  • Micro-stops
    the sub-minute events paper never captures
  • Comparable
    across lines, shifts and sites, on one definition
The problem

Why the OEE you have doesn't help you decide

  • It is declared, not measured

    If an operator writes the stops down at the end of the shift, the data reflects what could be remembered. Causes collapse into two or three generic buckets and durations round to the nearest five minutes.

  • Micro-stops vanish

    A forty-second stop gets recorded by nobody, but two hundred micro-stops in a shift cost more than one long breakdown. It is the item that erodes performance without ever appearing in the report.

  • Two identical lines aren't comparable

    If each department defines ideal cycle time its own way, the gap between two identical lines cannot be read. Without a shared definition, OEE is something you report, not something you act on.

The method

Three steps, from machine tag to units per week

Take state from the machine

Running, stopped, alarm, part count and scrap come from the PLC over OPC-UA, S7 or Modbus. No terminal to fill in, no shift to reconstruct from memory.

Settle on one definition

Ideal cycle time, stop reasons and the micro-stop threshold are defined once and hold for every line, so the numbers become comparable rather than merely correct.

Translate the gap into units

The difference between two lines or two shifts converts into units per week and euros of margin. That is the form in which a priority gets decided rather than debated.

What you can ask

Plain-language questions, answers from your own plants

  • What was line 2's OEE this week, and what pulled it down?
    The value broken into availability, performance and quality, with the top causes by lost time underneath. The component that weighs shows immediately, without opening a dashboard.
  • How many micro-stops did we have yesterday, and where do they cluster?
    Event counts below the threshold, grouped by machine and by hour. Usually one station and one specific point in the shift stand out.
  • Why does line 1 out-produce line 2 by 12% on the same product?
    A line-by-line comparison across changeover, actual speed, scrap and stops. The gap decomposes into the few items that actually generate it.
  • What does the packer's recurring stop cost us in units?
    Frequency times average duration, multiplied by ideal cadence, expressed in units per week and margin lost.
Why it holds up

A number that survives the meeting

  • Traceable to the tag. Any OEE figure can be opened down to the machine event that produced it. When somebody disputes the number, the answer is a timestamp, not an opinion.
  • No load on operators. There is no extra terminal to fill in at end of shift. The number the floor needs and the number management needs are the same number.
  • Complete history. Series stay at their original granularity, so a seasonal pattern or a slow drift is still recognisable twelve months later.

Let's start with a single line.

A guided pilot on the line that gives you the most trouble: within a few weeks you have real OEE, micro-stops counted, and the gap to its twin line expressed in units.

FAQ

Frequently asked questions

  • Do we need an MES to calculate OEE?

    No. OEE is computed from signals the PLC already exposes: machine state, part count and scrap. If an MES is present, Meddle reads it as another source and uses its stop reasons rather than inventing new ones.

  • How is a micro-stop defined?

    By a duration threshold you set, typically between thirty and sixty seconds. Below it the event is counted and attributed but not treated as a production stop, which keeps the statistics readable.

  • Does it work on machines that don't expose a part count?

    Yes, with a stated approximation. Where the count is missing it is derived from machine cycle or an added sensor, and the method used stays visible next to the figure so nobody mistakes an estimate for a measurement.

  • How much history is needed before the numbers are useful?

    A week of production is enough to see micro-stops and the gap between shifts. Seasonal comparison or slow drift needs a few months — which is why it pays to start collecting before you have decided what to measure.

  • Is OEE comparable across different sites?

    It becomes comparable once the definition is single. Ideal cycle time, reasons and thresholds are set at the model level and apply everywhere; that is exactly the step that makes the number useful beyond one department.

Let's look at the real OEE of your worst line.

In a guided demo we start from your own machine data, not from an example.

Book a demo