Roll out an MES in this order: baseline one line, write requirements, survey machines and signals, set up master data, connect a pilot, check it against a hand count, train and go live, fix the top loss, then expand line by line. Leave ERP integration and extra modules until the pilot's data is trusted.
Why the order matters
An MES project can do all the right things and still hurt, if it does them in the wrong order. Connect all lines before the reason codes are settled, and you get to redo that work on every line.
Each step below ends with an exit test. Do not start the next step until the test is passed.
Measure OEE and stops by hand.
Musts, shoulds, and a test for each.
What each machine can give you.
Lines, ideal cycles, reasons, shifts.
One line, full signal set.
Compare the system with a hand count.
Every shift, at the line.
Prove the data changes something.
Then integrate, then add modules.
Before you connect anything
1. Baseline one line
Choose a line that matters and a team that is willing. Measure two weeks of OEE and stops by hand, with the OEE tracking template.
Exit test: you can state availability, performance, quality and the top three losses in minutes.
2. Write the requirements
Turn that baseline into a list you can test, and shortlist against it. The MES requirements checklist gives you the lines, and light MES vs full MES helps you scope it.
Exit test: each must has a named test, and you have run the ROI sum at half the planned gain.
3. Survey machines and signals
Walk the pilot line with a controls technician. For each machine, write down what can provide state, count and speed, and how.
A modern controller may give all three. An older machine may need external sensors, and PLC signals for OEE lists what to ask for.
Give that table to each option on your shortlist, and make the final choice only after they have answered it. Otherwise you may buy something that cannot read your oldest machine.
Exit test: a one-page table of machine, signal, source and owner, with network access agreed with IT and the choice made.
4. Set up master data
This is the dull step that decides whether anyone trusts the numbers. Settle the shift calendar, the rules for breaks, the ideal cycle time per product and the reason codes.
Exit test: operators from each shift have read the reason list, and two of them code the same sample stops the same way.
Prove it on one line
5. Connect the pilot
Connect one line completely, including the awkward machine. A pilot that skips the hard machine proves little about the rest of the plant.
Exit test: state, count and speed reach the screen for a full shift, with no gaps you cannot explain.
6. Shadow run: check it against a hand count
Keep the old way running next to the new system for a while. Every difference is a bug in the signal, the master data or the old way, and it is cheaper to find now.
| Item, one shift | System | By hand | Difference |
|---|---|---|---|
| Total count | 5,212 | 5,190 | 0.4% |
| Rejects | 41 | 44 | 3 units |
| Stop minutes | 71 | 52 | 19 min |
| Number of stops | 23 | 9 | 14 stops |
Read each gap and decide which side is right. Here, the 14 extra stops average just under a minute and a half each (19 / 14), which fits micro-stops nobody writes down.
Agree your own tolerance before you start. Example: count within 1%, and every stop gap explained.
Exit test: the line's supervisor agrees the figures are right, and says so in front of the team.
7. Train and go live
Train every shift at the line, on their own stops, not in a meeting room. OEE training for operators has a lesson plan.
Explain what the data is for before anyone sees it, because operator buy-in is much harder to win back than to win.
Exit test: most stops above your threshold get a reason, on all shifts, without reminders.
8. Fix the top loss
This is the step that plans tend to forget. Use the new data to remove one loss, and show the result.
A short daily production meeting on yesterday's top stop is enough to get going. The OEE improvement action plan holds the actions.
Exit test: one loss has moved in the data, and the team knows why.
What the first line can look like
Example: a made-up plan for a first line, in weeks. Your machines, IT rules and purchasing will change it, and purchasing alone can take longer than this whole plan, so treat it as a shape and not a promise.
Scale it
9. Expand line by line
Copy what worked, and keep the master data rules plant-wide. Add lines in a sensible batch, with a shadow check on each new machine type.
After that, connect the ERP so orders and good counts flow, and add other functions as MES modules describes. By then you will know what you need, because your data will show it.
Exit test: line figures roll up correctly, and line and machine OEE mean the same thing in every report.
Order mistakes to avoid
- ERP integration first. It can take the whole project's energy, and it gives no view of stops. Do it once the pilot's data is trusted.
- Big-bang go-live. When every line starts at once, no one has time to check any of them.
- Reason codes written in the office. Operators will not use a list they had no part in, so test it at the line.
- No owner after go-live. Name one person who owns the data and the codes, before the project team moves on.
- Stopping at the screens. Live numbers with no meeting and no actions change nothing.
Step 5 is where mixed machines cause trouble. Fabrico is a light MES for OEE that reads stops, speed and counts live from PLCs, or from external IoT sensors on older machines, so a pilot line can include its awkward machine. See how Fabrico connects a pilot line.
What people ask next
How long does an MES implementation take?
It depends on the machines, the IT rules and how much of the system you take on. A single line with a light scope moves far faster than a plant-wide full MES. Map your own plan from the steps above, and use the pilot to reset it.
What is the first step in implementing an MES?
Measure one line by hand and state the loss you want to fix. You need that baseline to write requirements, and later to prove the gain. The OEE calculator is enough to begin.
How do I choose the MES pilot line?
Pick a line that matters, ideally a bottleneck, with a willing team and at least one awkward machine. An easy line gives an easy pilot and teaches you little.
When should the MES connect to the ERP?
Usually after the pilot's data is trusted, unless an order or traceability rule needs it sooner. MES vs ERP shows which data should move in which direction.
What is an MES shadow run?
A period when the new system runs next to your old method, and you compare the two on the same shifts. It catches wrong signals and bad master data before anyone relies on the numbers. Check gaps against downtime data accuracy.
Fabrico is a light MES for OEE. It takes stops, speed and counts from PLCs or IoT sensors, with add-on AI cameras counting output where a machine has no usable signal, and operators tag the cause at the line.