In short: OEE monitoring replaces shift-sheet guesses with machine signals, and in most UK plants it shows a lower, more useful figure than the clipboard did, because the short stops nobody writes down turn out to cost more than the breakdowns everybody remembers. The sensors for older machines are cheap; the harder job is getting a reliable, timestamped signal and an operator reason code back from every machine on the floor so connectivity is the first question to settle, ahead of the software.
Key Takeaways
- Hand-logged OEE flatters the plant — shift sheets capture the forty-minute breakdown but miss the dozens of two-minute jams, so the performance losses that cost most output never get a cause.
- Old machines can report too — a current clamp, a stack-light tap or a cheap I/O module gives a 1990s lathe the same run, idle and down signal as a new CNC with an OPC UA interface.
- Coverage at every machine is the hard part — the data volumes are tiny, but every press, operator tablet and edge gateway needs a link that stays up behind steel, which is where most OEE rollouts slip.
In a nutshell
What is OEE monitoring, and why do hand-logged figures mislead?
OEE monitoring is the automatic measurement of overall equipment effectiveness from machine signals, rather than from figures an operator writes on a shift sheet at the end of the day. OEE itself is the product of three ratios, availability, performance and quality, and the 85% figure that appears on so many factory noticeboards comes from Seiichi Nakajima's work on total productive maintenance in the 1980s, where it was framed as a world-class target. Vorne, which sells OEE software and publishes a great deal on the subject, puts typical discrete manufacturing nearer 60%.
The gap between those two numbers is where clipboards cause trouble. An operator reliably records the events that feel like downtime: the spindle that seized, the changeover that overran, the morning the material did not arrive. Operators rarely record the stop of ninety seconds to clear a jam, the five minutes waiting for a forklift, or the machine running at 80% of its rated speed because someone slowed it down after a quality scare in 2019 and nobody sped it up again. Those losses are real, they recur dozens of times a shift, and on paper they vanish into a vague "performance" number with no reason attached. A plant that believes it runs at 80% and measures 62% once the machines report for themselves has found where its output was going.
How do you calculate OEE?
OEE is availability multiplied by performance multiplied by quality, and a worked example shows why automatic measurement matters. Take an eight-hour shift of 480 minutes with a 30-minute planned break, leaving 450 minutes of planned production time.
- Availability: the machine was stopped for 60 minutes of unplanned downtime, so it ran for 390 minutes. Availability is 390 / 450 = 86.7%.
- Performance: the ideal cycle time is 30 seconds per part, so 390 minutes of running could have produced 780 parts. The machine made 624. Performance is 624 / 780 = 80%.
- Quality: 600 of the 624 parts were good. Quality is 600 / 624 = 96.2%.
OEE is 86.7% × 80% × 96.2% = 66.7%. As a check, 600 good parts at 30 seconds each is 300 minutes of fully productive time, and 300 / 450 is the same 66.7%.
The performance line is the one to look at. The machine lost 156 parts, or 78 minutes of output, while it was nominally running. A clipboard will usually record the 60 minutes of downtime with a cause next to it and leave the larger 78-minute loss with no explanation at all. Automatic machine downtime tracking splits that 78 minutes into micro-stops and slow running, timestamps each one, and lets the operator attach a reason while they still remember it.
What does machine downtime tracking need from each machine?
Every machine needs to provide three things: (i) its state, running, idle or down, with an accurate timestamp; (ii) a part count, ideally with a reject count; and (iii) a reason code whenever it stops for longer than a set threshold, entered by the person standing next to it.
Newer machines make the first two easy. A CNC from the last decade will usually expose state and counts over OPC UA, MTConnect or a PLC register that an edge gateway can poll. The difficulty is the rest of a typical UK machine shop, where a 2021 five-axis machine sits beside a 1994 lathe, a press with relay logic and a bought-in packing machine whose controller nobody has the password for. Those machines can still report. A current transformer clamped round the spindle motor feed tells you when the machine is cutting; a light sensor on the stack light gives you green, amber and red; a proximity sensor or a spare output on the ejector gives you a part count. These retrofit kits typically cost a few hundred pounds per machine and an electrician can fit them without touching the machine's control logic.
The reason code is the part that most often fails, and it fails for human reasons. If the operator has to walk to a shared terminal at the end of the aisle, they will batch their reasons at break time, and the batch will say "other". A tablet or small touchscreen mounted at each machine, prompting the moment a stop passes two minutes, gets real answers. That tablet is a network device, and it has to work at the machine itself, which is often the far end of the building from the office.
Why is connectivity the hard part on a shop floor?
The data from OEE monitoring is small. A state change, a count and a reason code amount to a few kilobytes per machine per hour. The problem is reach and reliability, and a typical shop floor makes both difficult.
Firstly, the machines are the obstruction. Press frames, CNC enclosures, racking and steel mezzanines block and reflect 2.4 GHz and 5 GHz signals, so a Wi-Fi survey done in an empty bay on a Sunday rarely matches the floor at 10am on a Tuesday. We have written about why warehouse and factory Wi-Fi fails in these conditions; OEE terminals sit in exactly the corners those surveys miss.
Secondly, gaps corrupt the record. An edge device that loses its link for twenty minutes and buffers the data is fine, provided it timestamps locally and back-fills cleanly. A tablet that drops off the network mid-shift stops collecting reasons, and the loss then appears as unexplained downtime, which is the very problem the system was bought to remove.
Thirdly, cables are expensive where the machines are. An Ethernet drop to a machine in the middle of the floor needs a route through cable tray, often overhead, often across a production area that has to stop while the work is done. Ten drops is a manageable cost; sixty drops for a whole shop, plus the relocations every time the layout changes, is a sizeable project in its own right.
Finally, the machine network should not share space with everything else. Cyber Essentials and IEC 62443 both push towards keeping operational technology segregated from office and guest traffic, and many plants' existing Wi-Fi carries visitor laptops and personal phones on the same access points the OEE tablets would use.
When is private 5G worth it for OEE monitoring, and when is it not?
Private 5G is worth it when OEE monitoring is one of several workloads that all need reliable coverage across the floor. It is not worth it for OEE alone on a small site, and we would say so to any customer who asked.
For a unit with a dozen machines along one wall, a couple of well-placed industrial Wi-Fi access points on a segregated SSID, or a few Ethernet drops, will carry OEE data perfectly well at a fraction of the cost. LoRaWAN is another sensible option for part counts and state signals if operator tablets are not needed.
The calculation changes as the plant grows and the list of connected things lengthens. A site with 60 or more machines, mobile operator tablets, handheld scanners, AGVs, condition monitoring sensors and perhaps vision inspection cameras has a coverage and capacity problem that one licensed network handles well. Private 5G runs on spectrum that Ofcom licenses to the site under its shared access scheme, so there is no interference from neighbouring units or visitors' hotspots. It gives each device its own SIM identity, which makes OT segregation a matter of configuration rather than physical separation, and it covers a large shop from a handful of radios mounted high, avoiding a cable run to every machine. The Wi-Fi 6 versus private 5G decision sets out the trade-offs in more detail. For a plant in that position, OEE monitoring is usually the cheapest application to add to a network it already needs, and one of the first to show a return.
Can Made Smarter funding pay for OEE monitoring?
In England it often can, in part. The Made Smarter adoption programme offers small and medium-sized manufacturers a funded digital roadmap and, in most regions, match funding of up to 50% towards hardware and software, typically capped at £20,000, with a limited number of larger grants of up to £50,000 in the South East. OEE monitoring suits the programme well, because the business case is easy to state and the results show within a quarter.
The details vary by region and by year, so a plant should check the current terms on the Made Smarter site for its own area before planning around a grant. Scotland, Wales and Northern Ireland have separate support routes through Scottish Enterprise, Business Wales and Invest NI.
Where should a plant start?
Start with the bottleneck machine, the one whose output sets the pace for the rest of the line, and measure it automatically for four weeks before changing anything. Plants that do this usually find two or three recurring short stops that account for most of the lost performance, and fixing those pays for the system.
From there the rollout follows the money: the next constraint, then the cell around it, then the floor. At each step, check the connectivity before buying more sensors. Walk the floor with a phone or a survey tool during a busy shift, note where the tablets would sit and where the signal drops, and decide early whether the plant is solving one OEE problem or building the network for everything it plans to connect over the next five years. If it is the second, talk to us about a coverage survey for the manufacturing floor before the first retrofit kit goes on.