PlatformMonitoring & OEE
Know — monitoring, OEE & dashboards
Production monitoring: which machine is slowing the shift.
Production monitoring software is only worth having if you would put its numbers in front of a customer. Every state, cycle, stop and scrap event is measured at the machine and lands in one ledger — so the number on the wall, the number in the report and the number in the audit are the same number.
Live machine state
The true state of every machine, derived rather than declared.
Production line monitoring starts with knowing what each machine is doing right now, and no one types that into SmartFactory. The system derives every state from cycle data, power draw and device liveness, and changes it on its own.
- Seven states you can act on — No Job, Setup, Running, Running Slow, Running Fast, Stopped and Offline — each derived from real cycles rather than a whiteboard magnet.
- A machine counts as Stopped against its own rhythm: the threshold is derived from that machine’s standard cycle time, so a 20-second press and a seven-minute press are judged on their own terms.
- Running Slow and Running Fast trigger on a configurable band around the standard cycle, evaluated over the most recent cycles rather than one unlucky shot.
- Offline is a state, not a silence: if a machine stops reporting, the board says Offline instead of quietly holding the last good number.
- Every state change is committed to a permanent ledger with its timestamp — the evidence behind every OEE figure you will ever be asked to defend.
This is the layer a production monitoring system stands or falls on. If the state ledger is wrong, every number built on top of it is wrong too, however good the dashboard looks.
OEE & works orders
An OEE figure that survives the meeting.
Most OEE arguments are really arguments about the inputs rather than about the arithmetic — which is how two OEE systems can look at one shift and disagree. SmartFactory shows its working, and reports the awkward version of every number alongside the flattering one. If you are still choosing, what to ask a vendor covers the questions worth sending first.
- Availability comes from the machine-state ledger against loading time — your real shifts, minus the downtime you have classified as planned. Change the working calendar and history re-states itself.
- Performance is reported raw and capped at 100%: if a machine is beating its standard cycle, you see it rather than having it quietly rounded away.
- Quality comes from real inspection results and scrap events, not an assumed yield.
- Works orders carry live parts-made counts from cycles × cavities, remaining quantity, and a completion forecast banded On Track, Behind or Critical at 95% and 80% attainment.
- Reliability falls out of the same ledger: MTBF and MTTR per machine, so “which machine breaks most often” and “how long does it take us to get it back” are two different questions with two different answers.
- If a machine has no shift pattern defined, SmartFactory says so on the tile rather than inventing planned time.
Downtime
Every stop opens its own event — before anyone reaches for a clipboard.
- A machine stopped past its grace period opens a downtime event automatically, timestamped from when it actually stopped rather than when someone noticed.
- Operators classify it in two taps on the machine-side tablet, against your own two-level reason tree — breakdown, changeover, material, quality, planned, other — with a note demanded where the reason is “other”.
- Downtime Pareto ranks what is really stopping you, and scrap-by-reason does the same for the parts you threw away.
- An unclassified-downtime figure sits alongside it, telling you what share of your downtime story is currently fiction.
- Micro-stops shorter than the grace period never invent a downtime event, but they stay in the state ledger — so the cycle-anomaly view can still find the machine that stalls twenty times a shift.
Dashboards
Five roles, five boards, and one written question per tile.
Nobody wants to build a dashboard. Every preset ships answering the questions that role actually asks — and stays customisable once it is yours.
- Role presets for the factory overview, production, maintenance, quality and energy — each with a default time window that suits the job, from shift-to-date to the last 30 days.
- Every tile is registered with the question it answers, in writing: “Which machine is dragging the shift down?”, “What is stopping us, ranked, and is it getting worse?”, “How much of our downtime story is fiction?”
- A four-level drill that always goes the same way: factory → location → machine → the event or job behind the number. No dead ends, no going back to the top.
- Your time window and filters live in the URL, so a board survives a refresh, and a link you paste into an email opens on the same evidence you were looking at.
- Installs like an app on any tablet from the browser — no app store, no rollout — with push notifications for the alerts that matter. Phones are supported too, for the people you want reachable away from the floor.
Andon TVs
A screen on the wall that can’t lie, and doesn’t need a login left on it.
- Pair a display by scanning a QR code from the admin screen — the TV gets its own read-only display token, not a person’s account left signed in on a corridor screen.
- That token can see the boards and nothing else: no operator station, no exports, no writes. Revoke it from the admin screen and the screen goes dark on its own.
- States are readable by everyone: colour plus a glyph plus the word, because roughly one in twelve men on a shop floor cannot reliably separate the red and green andon boards are built from.
- The screen keeps itself awake using the browser’s wake-lock, and carries a live connection pip with a “last updated” stamp — so a frozen board announces that it is frozen instead of showing yesterday’s good news.
- Auto-rotation walks machine pages on a set interval, and keeps its place even when a machine drops off the board mid-cycle.
Honest dashboards
The MES that tells you when it doesn’t know.
This is the answer to the question that kills MES deals: can I trust the number? Monitoring software is trusted or quietly ignored on that one property, so every board here is built to fail visibly rather than plausibly.
- An empty window says “No production in this window” — never a convenient 0% OEE that someone screenshots into a board pack.
- A figure affected by a sensor dropout is shown with the caveat attached, in amber, next to the number rather than in a footnote nobody reads.
- A tile nobody has configured links you to the setting that fixes it instead of printing a zero.
- Where two honest numbers exist, you get both: raw and capped performance, schedule-late and forecast-late, downtime MTTR and repair MTTR, measured machine state and reported fault state.
- Forecasts built on thin data are visibly muted, so a confident-looking line is always a line with evidence behind it.
Alerts & reports
The right numbers, in the right inbox, before you ask.
- Alerts on the four things that actually cost you money: a machine stopped too long, a device gone offline, energy spiking over baseline, and a job falling behind schedule — in-app, by email and by push.
- The energy-spike rule compares against a rolling baseline for the same hour of the same weekday, and refuses to make a call until it has enough history — so Monday morning start-up never pages anybody.
- Cooldowns and hysteresis are built in: alerts clear on their own and do not re-fire every five minutes while you are already dealing with it.
- Scheduled reports — OEE by machine, downtime Pareto, quality summary, scrap by reason, job history, energy and machine cycles — arrive as branded PDF, CSV or Excel on your schedule, in your timezone.
- Everything on screen exports, so the analysis you insist on doing your own way still starts from our data.
See your factory live in thirty minutes.
Book a demo and we’ll walk you through a running factory — or register for a free trial and get a simulated factory of your own to play with, the same day. No hardware, no card, no IT project.