"We know we lose time on changeovers, but nobody knows how much." That one sentence, repeated across a dozen conversations with owners of small Polish factories in the last year, sums up the state of production measurement in SMEs. Large companies have MES systems worth €90k. Small ones have the foreman's notebook and the owner's intuition. Between those two things sits a huge space where most of the improvement doesn't require an MES at all — it just requires someone to finally start measuring.
MES-lite isn't a product. It's a way of thinking: collect the minimum data you need to make a good decision, and nothing more. You don't build a data warehouse on day one, you don't buy an APS module, you don't integrate six systems. You put a tablet next to one machine, the operator taps "start / stop / reason for stoppage", and after a month you see for the first time in your life where the hours actually disappear to.
This post is about what such a project looks like in practice — what's enough to measure, what it costs, and what you absolutely should not do at the start.
What you don't actually know about your own shop floor
Before you buy anything, it's worth honestly answering three questions about your production. If any answer is "well… roughly", you have a problem that MES-lite solves in a few weeks.
- How many hours a week do your machines stand still, and why — broken down by changeovers, breakdowns, missing material, missing operator, no work orders?
- What is the real OEE (availability × performance × quality) on your key machine? Not "roughly", an actual number.
- How many pieces of a given product come off the line per hour in a normal rhythm, and how many in a bad week?
Owners typically believe their OEE is around 75-80%. In real measurement in the small metalworking, plastics, and welding shops we've audited, it comes out at 45-60%. Not because something is broken — because nobody was measuring, and "a standing machine is normal, it'll start again in a minute". The first month of measurement is always a reality check.
A concrete example: 6 CNCs and a €200 tablet
At one of our clients — a 25-person company in south-eastern Poland machining steel components for the machinery industry — the shop had 6 CNC machines (mills and lathes) running two shifts. The owner assumed OEE was around 70%. He was about to invest in a seventh machine.
We built a MES-lite in 4 weeks. Next to each machine sat a €200 Android tablet in a simple protective enclosure, running a small web app (Vue frontend + REST API to a PostgreSQL database on a €10/month VPS). The screen had three buttons: "Start operation", "Stop — pick a reason", "End work order". The stoppage reason list was cut down with the foremen to 8 items (changeover, no material, mechanical breakdown, electrical breakdown, no operator, no program, quality check, other). Work orders were pulled from Comarch Optima via an hourly API call.
After a month of data — first surprise: average availability was 61%, not 80%. Second surprise: 3.2 hours per shift per machine was being lost to "no operator". It turned out the foreman was honestly moving people between machines, but the work-order schedule in Optima was built without any awareness of machine load — 60% of the orders needed the same operator with certification on one specific lathe type.
Rearranging the schedule and cross-training two more people on the second lathe type produced a +14% shop-floor throughput in three months. Without buying the seventh machine (planned cost: €85k). The full MES-lite project cost: €4,900 one-off (6 tablets, enclosures, application, Optima integration, shop-floor deployment) plus €35/month (hosting, licences).
This isn't a story about technology. It's a story about the fact that if you don't measure, you guess — and guessing in production costs €85k for a machine you don't need.
What to measure first — and what to postpone
The biggest SME mistake with MES: they want to measure everything at once. Result: the project balloons to €45k, deployment takes 8 months, operators hate the tablet, and the owner ends up staring at 40 dashboards nobody has time for.
Start from the minimum. On the first machine (or a group of 3-6 similar ones) you measure only three things:
- Running time vs. idle time — when the machine is actually working, when it's standing.
- Reason for each stoppage longer than X minutes (usually 5-10) — from a short list of 6-10 categories built with the foreman, not by a consultant behind a desk.
- Pieces / metres / kilograms produced at end of operation (if the machine doesn't count itself, the operator enters it manually).
That's it. Not process parameters, not temperatures, not power draw, not sub-minute microstops. You'll add those in the next phase, if you can prove they matter. In phase one, the point is that management sees the real picture for a month — no cosmetics, no estimates, no "well, we all know that's how it is".
Hardware and budget — realistically
You don't need industrial operator panels costing €1,800 each from engineers in suits. A typical SME MES-lite fits in this basket:
- Android tablet (Samsung Galaxy Tab A9 or similar) for €150-270, in a protective case for €35, screwed to the machine or on a VESA mount.
- Power run from the nearest 230V socket by an electrician (one-off, €45-90 per machine).
- Shop-floor Wi-Fi — if you already have a decent access point, that's enough. If not, invest (you can't run MES on the single router from your ISP).
- Web application — can be simple, sensibly built for €3,500-7,000 up-front if it only does the 3 measurements. Off-the-shelf MES products aimed at small manufacturers run €180-550/month — compare against building your own.
- Integration with ERP work orders (Optima, Subiekt, enova) — if the ERP exposes an API, that's another €1,200-3,500. If it doesn't, you start by entering the work-order number by hand, and that's also fine for phase one.
Realistic range for a shop with 4-8 machines: €4,500-11,000 one-off, plus €45-110 per month. That's a figure where the payoff arrives after the first well-informed decision about throughput, and often after the first avoided investment in an unnecessary machine.
What not to do at the start
Three traps SMEs regularly fall into when "someone at a trade show sold them on an MES":
Don't buy an off-the-shelf MES without a pilot. Vendor demos show 40 modules: scheduling, traceability, quality, TPM, energy, SCADA integration. You buy it and spend a year rolling out something 80% of which you'll never use. Start with a pilot on one machine in four weeks — from a pre-built product or your own. Only after three months of real data do you know what you actually need.
Don't measure micro-stops until you've won on macro-stops. Recording 10-second stoppages from the PLC sounds professional, but if you're losing 3 hours per shift to "no operator", fighting for 30 seconds is a waste of your attention. Fix the big things first.
Don't put a dashboard in the owner's office until you have the foreman on your side. MES without shop-floor buy-in is a source of conflict. Operators start "gaming" the taps, data becomes garbage, the project dies quietly. Run the first week of the pilot with the foreman, not with the board. Without the foreman, MES is snitching; with the foreman, it's a work tool.
What next
If you have production in your company and you suspect OEE isn't where you think it is, the first step costs one conversation and a week of watching one machine with a tablet and a notebook. You don't need to buy a system to measure it that first time. You just need to decide you're going to stop guessing.
We help SMEs put projects like this together, from a pilot to full ERP integration. The first call is free and takes half an hour — enough to tell whether MES-lite makes sense in your case, or whether you should start from a completely different angle.
A small production shop doesn't need a half-million-euro MES. It needs to know where the hours disappear. Everything else follows from that one answer.