A clean tracker any plant can adopt, and a plain warning for whoever signs off on the number: the OEE figure rebuilt in a spreadsheet days later is almost always higher than the real one. If capital decisions ride on that number, the way it is collected is a board-level question, not a floor detail.
Read this first · Why this is your problem, not just the floor's
OEE is a single number that stands in for how much of the plant you paid for is actually turning into good product. It gets quoted in reviews, put in front of boards, and used to justify or defer capital. So the quality of that number is an executive concern. A figure that looks respectable because the losses were never captured will steer you toward the wrong spend.
There are no invented numbers on this page. Every figure below is one you measure on your own lines. We do not tell you what a good score is or what other plants run, because an average from someone else's floor cannot price a decision on yours.
What OEE actually measures
OEE, Overall Equipment Effectiveness, is three ratios multiplied together. Each one is a different way the paid-for capacity of a line leaks away. Multiplied, they give one figure between zero and one hundred percent for how much of the possible good output you actually got. The three components are the free part of this page, because you cannot judge a tracker without them.
Component 01
Availability
Run time ÷ planned production time
Of the time you planned to run, how much the line was actually running. This is where breakdowns, waiting on material, and changeovers land. It is the loss people remember, because a stopped line is obvious.
Component 02
Performance
Actual output ÷ output at ideal rate
While it was running, how close the line ran to its ideal speed. This is where slow cycles and small stops of a few seconds hide. It is the loss no one remembers, because nothing looked broken.
Component 03
Quality
Good units ÷ total units
Of what the line made, how much was sellable the first time. Scrap and rework live here. Product that had to be reworked cost you the capacity twice, even if it eventually shipped.
OEE is Availability times Performance times Quality. Because the three multiply, a plant can run each component in the high eighties and still land an OEE in the sixties. That compounding is the reason OEE is worth tracking at all: it exposes losses that any single metric would let you rationalize away.
Why the Excel-days-later version misleads
Most plants already have an OEE number. The problem is where it comes from. Operators jot stops and counts on a sheet during the shift, and someone rebuilds it in a spreadsheet a day or a week later. That version is not just late. It is biased upward, and it is biased upward in a way that hides your largest recoverable loss.
Three failures compound in a hand-built tracker, and each one pushes the number in the flattering direction:
The micro-stops never get written down. A person cannot log a four-second stop or a cycle that ran ten percent slow while also running the line. So Performance loss, usually the largest and most recoverable bucket, is the loss most likely to be missing entirely.
The denominator drifts. Planned production time gets adjusted after the fact to explain a rough shift, which quietly inflates Availability.
One person owns the sheet. The number reflects their memory, their rounding, and their judgment on a bad day, and no system can read it, audit it, or feed anything from it.
The result is a number that looks defensible and is quietly optimistic. You cannot see this by staring at the spreadsheet, because the losses that are missing left no trace in it. The honest way to know how far off you are is to compare a hand-built figure against the same line measured from its own machine data. That gap is the free capacity you have been managing around.
What the gap is worth to the P&L
This is the part that belongs in your seat, not the floor's. A soft OEE number is not a reporting nuisance. It changes what you spend and where.
Capacity
Capex you may not need
Hidden Performance loss is unbuilt capacity you already own. Recover it and the line you were about to buy, or the shift you were about to add, may already be sitting inside the current asset.
Margin
Giveaway and rework
Quality loss is margin walking out the door as scrap and second-pass labor. A tracker that smooths it over lets it keep running because no one is holding a number that indicts it.
Decisions
A number you can bet on
A figure assembled from memory is not something to base a capital plan on. A figure computed from machine data is. The difference is whether the number can carry a decision.
You do not need us to put a dollar figure on your own lines. That is exactly what the OEE calculator does: enter your planned time, downtime, counts, and rejects, and it returns your Availability, Performance, Quality, and OEE with no sign-up. Run it once with your hand-tracked figures, then again once you can pull the same fields from the machine, and the honest gap is the size of the prize.
Should you spend on AI for this at all
The fair answer is: not first, and not as a product you bolt on. AI cannot improve a number it cannot see, and a hand-built tracker gives it nothing to see. The spend that pays is the unglamorous one underneath: getting the counts, states, and rejects off the machine and into one place that is current and entered once. Do that, and OEE becomes trustworthy on its own. It is also the exact foundation any later AI, scheduling, or forecasting would stand on, so the money is not a detour. It is the first phase of the same road.
The full template
Get the full tracker template.
You have seen the three components and why the days-later version overstates the number. Enter your work email and the full field-by-field tracker template opens right here on this page, along with exactly how to feed each field from machine data instead of by hand. A copy goes to your inbox.
The complete tracker template: every field, its formula, and the source it should come from
How to feed Availability, Performance, and Quality from the machine rather than a clipboard
The five steps to a tracker leadership can put in front of a board
Work email only. We use it to send the template and nothing else you did not ask for. Unsubscribe anytime.
Unlocked. The full template is open below, and a copy is on its way to your inbox. If you checked the box, a Harmony engineer will reach out to set it up against your machine data.
Template · Part 1
The tracker, field by field
One row per line per shift. The top block is what you capture. The shaded block is what the sheet computes from it. Nothing here is entered twice.
Field
What it captures
Where it should come from
Line / asset
Which machine or line this row is for.
Fixed list, one row per asset per shift.
Shift / date
The window this row covers.
Shift calendar.
Planned production time
Minutes you intended to run, after planned breaks.
Schedule. Agreed once and held steady, not adjusted after the shift.
Downtime
Minutes stopped when you meant to be running, including small stops.
Machine states and faults. This is the field a person cannot capture completely by hand.
Changeover / setup
Minutes lost to setup and changeover.
Machine state, tagged as changeover.
Ideal cycle rate
The fastest sustainable rate the asset is rated for.
Documented once per asset. The most argued-over field, so pin it down.
Total units
Everything the line produced, good or not.
Machine counter, not a clicker.
Good units
Units sellable with no rework.
Reject or inspection point, so scrap is not silently counted as good.
Availability
Run time ÷ Planned time
Computed. Run time is planned time minus downtime and changeover.
Performance
(Total units ÷ Run time) ÷ Ideal rate
Computed. This is where slow-cycle loss finally shows.
Quality
Good units ÷ Total units
Computed.
OEE
Availability × Performance × Quality
Computed. The one number, and the three you drill into when it drops.
Why the three computed rows stay split out
A single OEE percentage tells you something is wrong but not what. Keeping Availability, Performance, and Quality visible on every row is what turns the tracker from a scorecard into a diagnosis. When OEE drops, one of the three moved, and you know within seconds whether to look at stops, speed, or scrap. Roll them up only for the board slide, never for the people who can act on them.
Template · Part 2
How to feed it from the machine, not the clipboard
The template is only as honest as its inputs. Here is where each captured field should come from once you stop hand-entering it, and what changes when you do.
Field
By hand today
From machine data
Downtime
Written down when someone notices and has a free hand. Short stops missed.
Read from the controller's run and stop states, so every stop counts, including the seconds-long ones.
Total units
Read off an HMI screen or a clicker at shift end and transcribed.
Pulled straight from the machine counter, continuously, with no transcription step.
Cycle rate
Assumed from memory of what the line usually does.
Measured against the documented ideal rate, so Performance loss becomes a real number.
Good vs reject
Reconstructed later from scrap tickets or estimated.
Captured at the reject or inspection point as it happens.
The mechanism, in plain terms
Most modern controllers already know their own counts, states, and fault codes. On many plants those numbers stop at the panel and never leave. A standard connection such as OPC UA lets the counts and states come off the PLC and into the tracker without a person retyping them. The point is not new sensors. It is getting the measurements the machine already has to stop dying at the panel.
When the feed is live, the tracker stops being a rear-view mirror. A bad run is visible during the run, not at tomorrow's meeting, which is the only version of OEE that can actually change a shift while it is still happening.
Template · Part 3
The five steps to a number a board can trust
In order. Skipping the early ones is why so many OEE programs produce a number no one believes.
Fix the denominators first. Agree what planned production time means and record it the same way every shift. Availability is meaningless if the window keeps moving.
Capture Availability from machine state, not memory. Take run and stop time from the machine so every micro-stop counts.
Capture Performance against a documented ideal rate. This is where the largest recoverable loss finally becomes visible.
Capture Quality at the reject point. Record good units and rejects where they happen, so rework lands in the number.
Multiply, and refresh in near real time. Compute OEE from live fields so the number can change a shift, not just describe it later.
Put your own numbers through it
Once you have the fields, price them. The OEE calculator takes planned time, downtime, counts, and rejects and returns your Availability, Performance, Quality, and OEE, free and with no sign-up. Run your hand-tracked figures and your machine-fed figures side by side, and the gap between them is the capacity the old tracker was hiding.
Where a real OEE feed sits: the three phases
An OEE number you can bet on is not a bolt-on. It is the payoff of the first phase every plant moves through, in order. The tracker you just unlocked lives in Phase 1, because a number is only as good as the digitized, connected, unified data underneath it.
Phase 1
Lay the Data Foundation · Digitization
Every pen-and-paper record digitized at the station, every software system connected, and all of the data unified into one live layer. Trustworthy OEE starts here.
Phase 2
Production & Operations Scale
Factory operations turn proactive: live sensors and machine data, the AI scheduling board, predictive maintenance before failure.
Phase 3
AI-Native Operations
Agents across the floor and the back office act on the live layer: quality signals, reports, copilots. Humans approve.
Want to see where your own floor sits before you spend on any of this? The AI Readiness Checklist walks the same foundation station by station, and the OEE calculator prices the gap on your own inputs.
Rather have this built against your real machine data?
Standing up a live OEE feed is the first week of a Harmony pilot, on-site, with forward-deployed engineers wiring the tracker to your own controllers. A fixed pilot is $15,000 to $20,000 one time and runs 4 to 6 weeks, with working software in your hands by the end of the pilot. See what the live layer looks like.
