Measure OEE First: Why MES Can Wait
Software vendors, us included, will happily sell a factory a manufacturing execution system. Here is the uncomfortable question to ask first: does anyone in the building know your OEE? If nobody can say what fraction of planned time the bottleneck line converts into good product, the MES conversation is premature. Measure OEE first: the number shows where the losses are, and the losses tell you what the software actually needs to fix.
The Maturity Ladder From Paper to MES
Factories climb the same ladder in the same order, and it pays to know your rung:
- Paper: what happened lives in logbooks, whiteboards, and the memory of the senior operator.
- Excel: someone types shift summaries into a workbook: hours old, rounded, only as good as the typing.
- Dashboard: live counts on an andon or SCADA screen: you know how many, not how much you lost or why.
- OEE: losses quantified: availability, performance, and quality against a planned rate, so the biggest loss has a number.
- MES: a system that acts on the losses: dispatches work orders, ties output to material lots, records reasons and checks.
Each rung funds the next. Skip the OEE rung and a MES lands on assumptions: nobody agrees on the baseline, so every improvement claim is arguable.
The OEE Math, Plainly
OEE is three percentages multiplied together:
OEE = Availability × Performance × Quality
- Availability = run time ÷ planned production time, where planned time is the shift minus planned stops like breaks and maintenance.
- Performance = actual output ÷ possible output at the ideal rate during run time.
- Quality = good units ÷ total units produced.
A worked example, one 8-hour shift on one line. Planned production time is 480 minutes. Downtime totals 60 minutes: a 35 minute changeover and a 25 minute jam, so run time is 420 minutes and Availability is 420 ÷ 480 = 87.5 percent.
The ideal rate is 60 units per hour, so run time allowed for 420 units; the line produced 357, and Performance is 357 ÷ 420 = 85 percent. Of those, 340 passed inspection, so Quality is 340 ÷ 357 = 95.2 percent.
Multiply: 0.875 × 0.85 × 0.952 = 70.8 percent. Respectable, until you see what hides inside: an hour of downtime and 17 scrap units every shift. The often-quoted world class figure is around 85 percent, and a factory's first honest OEE number is usually a surprise. That surprise is the map.
The Data You Need to Start
Only two things, so there is no excuse to wait for software:
- Machine states with timestamps: running, idle, down, changeover, plus a short agreed list of reasons for everything else.
- Counts: total units produced and scrap, per shift, per machine.
Automated lines get this from PLC counters and keyed SCADA ingest. A manual or semi-manual line gets it from a shift sheet with start and stop times and reasons. Imperfect data collected for a week beats perfect data collected never.
Start With One Line
Pick the bottleneck line, where an extra hour is worth the most. Define four or five states and agree the reason list with the supervisors who will use it. Record for one week, by hand on a clipboard if you must: the formula fits in three spreadsheet cells. Compute the three factors, find the biggest loss, fix that one thing, and measure again. Two weeks, almost zero cost, and afterwards any vendor demo has to answer one question: which of my measured losses does your system actually shorten?
Where Voltrus Fits In
Voltrus collects machine states through keyed SCADA ingest and rolls OEE up hourly, per station and per line, alongside downtime events with their reasons. That removes the tallying work. It does not replace knowing your line: compute it by hand for a week first, and you will know which losses a MES should automate and which just need a better changeover procedure.
For where a MES sits in the rest of the stack, see our comparison of MES versus ERP or the primer on what a MES is. If scheduling still lives in a workbook, our take on MES versus Excel scheduling covers that rung directly.
Frequently Asked Questions
What is a good OEE score?
Eighty-five percent is the commonly quoted world class figure, but chasing it without a trusted baseline is backwards. A 60 percent line with known reasons beats a guessed 90. Fix the biggest loss, then compare against your own last month, not against a poster.
Does OEE make sense for manual or packaging lines?
Yes. The operator logs the states instead of the PLC, the ideal rate comes from the work standard, and quality is whatever the pack check counts. Less precise than machine-fed data, but precise enough to rank your losses, which is all the first measurement is for.
We already show counts on a dashboard. Is that not OEE?
No. Counts are output; OEE needs time: machine states, reasons, and an ideal rate. A dashboard answers how many. OEE answers how much you lost, where, and why. The second question is the one worth paying for.
Measure First, Then Automate
Voltrus MES is live: one line to start, hourly OEE per station and line from keyed machine states. Bring your baseline and see what ships today.
See Voltrus MES