What are the loss categories in an OEE waterfall?

Four, always in the same order: schedule loss (time you chose not to run), availability loss (stops during planned time like breakdowns and changeovers), performance loss (slow cycles and micro-stops), and quality loss (scrap, rework, and startup rejects). Only the last three are inside OEE; schedule loss sits above the OEE line and belongs to TEEP and capacity planning.

How do you calculate the OEE waterfall steps?

Start with all calendar hours in your window. Subtract schedule loss to get planned production time, subtract availability loss to get run time, subtract performance loss to get net run time, and subtract quality loss to get fully productive time. Express each loss in hours by converting slow cycles and rejected units into the time they consumed at the ideal rate.

Is the OEE waterfall the same as TEEP?

They come from the same chart. OEE is fully productive time divided by planned production time; TEEP is the same fully productive time divided by all calendar time. The waterfall shows both because it keeps every block visible, OEE measures from the planned-time line, TEEP measures from the top.

Which loss should I attack first on the waterfall?

Read by size, then by recoverability. Availability loss is usually the biggest OEE bucket and the most actionable because stops have nameable causes. But if schedule loss above the OEE line is large, that block governs capacity more than any OEE fix, and it is a demand or staffing question rather than an efficiency one.

Why does my waterfall understate performance loss?

Because short stops go uncounted. When 90-second jams and micro-stalls are not logged, they never show up as availability loss and instead quietly reduce output that the performance factor should have flagged. Capturing stops at the source, from machine signals or operator tags at the station, is what keeps the small drains from disappearing into the big ones.

How often should you rebuild the OEE waterfall?

Build it once as a repeatable calculation and rerun it every shift, day, or week on the same window and the same definitions. The value is in the trend and the prioritized loss list, not a one-off slide. A stable schedule boundary and source-captured stops are what let you compare one week's waterfall to the next.

The OEE Loss Waterfall: Where Every Hour Goes

Q: What is an OEE loss waterfall?

It is a top-to-bottom accounting of production time that starts at all calendar time and subtracts one loss category at a time, schedule loss, then availability, performance, and quality loss, until only fully productive time remains. It turns the single OEE percentage into a stack of hours you can point at, price, and assign to an owner.

Reading OEE as a top-down time waterfall, draining calendar time through schedule, availability, performance, and quality loss until only good production remains.

An OEE loss waterfall is a top-to-bottom accounting of production time that starts at all calendar time and subtracts one loss category at a time, schedule loss, availability loss, performance loss, and quality loss, until only fully productive time remains. Each step shows exactly where an hour went.

The single OEE percentage tells you how much of your planned time turned into good product. It does not tell you where the rest went. The waterfall does. By draining time one bucket at a time, it turns an abstract score into a stack of hours you can point at, price, and assign. This post shows how to build the waterfall, what belongs in each step, and how to read it so the biggest recoverable hours rise to the top of the list.

What is an OEE loss waterfall?

It is a time waterfall: a chart that begins with the largest block of time and steps down through each loss until it reaches good production. Read the losses top to bottom and you have named every hour on the asset, the productive ones and every category that ate the rest.

The waterfall is the visual form of the same nested time model behind the OEE calculation. OEE compresses that model into one number; the waterfall keeps it stretched out so each loss stays visible. It answers a different question than the score does. OEE asks "how effective were we?" The waterfall asks "where did the time go, and which chunk is worth chasing first?"

That difference matters in a room. Hand a plant manager "OEE was 69% last week" and the honest response is a shrug, 69% of what, lost to what? Hand the same manager a waterfall and the conversation changes: fifteen hours gone to stops, thirteen to slow running, three to scrap, and sixty-eight above the line to a shift the plant never staffed. Now every block has an owner and a number, and the meeting can argue about the right thing. A percentage grades the past; a waterfall assigns the future. That is why plants that adopt the waterfall tend to stop debating the score and start debating the blocks.

Each step drains one loss category. OEE is the final productive block divided by planned time; TEEP would divide it by all calendar time.

What are the steps from calendar time to fully productive time?

There are four drains between all time and good product, and they always come in the same order. Naming them in order is what keeps a loss from being counted twice.

All time → Planned production time (minus schedule loss). Schedule loss is time you chose not to run: unstaffed shifts, weekends, no-demand idling, planned holidays. It is not an OEE loss, it lives above the OEE line, but it is the loss that TEEP and capacity planning care about most.
Planned time → Run time (minus availability loss). Availability loss is every stop during planned time: breakdowns, changeovers, material outs, and waiting. This is usually the fattest OEE bucket and the one your downtime log is built to catch.
Run time → Net run time (minus performance loss). Performance loss is speed: running below the ideal cycle, plus the small stops and micro-stalls too short to log as downtime. It hides well, which is why it is the bucket most often understated.
Net run time → Fully productive time (minus quality loss). Quality loss is time spent making units you cannot ship, scrap and rework, including startup rejects before the process stabilizes. Measured in time, not just count.

Fully productive time divided by planned production time is OEE. The same fully productive block divided by all calendar time is TEEP. One waterfall yields both numbers; the only difference is which line you measure from.

The fixed order is not cosmetic, it prevents double counting. Because availability is subtracted before performance, a machine that is stopped cannot also be charged with running slow; the stopped time already left the pool. Because quality comes last, a rejected unit is counted once, as the time it took to make at full speed, not smeared across performance and quality both. When a plant's numbers do not reconcile, the cause is almost always a bucket applied out of order or a loss claimed twice. Keep the sequence rigid and the arithmetic closes: the four drains plus fully productive time always sum back to planned production time, and planned plus schedule loss always sums back to all time.

How do you build an OEE loss waterfall?

Build it once as a repeatable calculation, not a one-off slide. Six steps take a line from raw logs to a waterfall you can rerun every week.

Fix the observation window and pull all time. Pick a period, a shift, a day, a week, and start with the full calendar hours in it. Every later number is a share of this block.
Subtract schedule loss to get planned production time. Write down, once, what counts as "not scheduled." This boundary is where most waterfalls silently disagree, so document it and keep it stable.
Subtract availability loss to get run time. Sum every stop inside planned time from the stop log. Captured at the source, not reconstructed from memory, which reliably undercounts short stops.
Subtract performance loss to get net run time. Compare actual output to what the ideal cycle would have produced in the run time. The difference is speed and micro-stop loss, expressed in hours.
Subtract quality loss to get fully productive time. Convert rejected and reworked units into the time it took to make them at the ideal rate.
Divide and label. Fully productive over planned is OEE; the four drained blocks are your prioritized loss list. Price each block in units or margin so the chart argues for itself.

How do you read the waterfall to set priorities?

Read it by size, then by recoverability. The tallest drain is the loudest, but not always the one to attack first, some hours are cheaper to win back than others.

Availability loss is usually the biggest OEE bucket and the most actionable, because stops have causes you can name and remove. That is where the six big losses framework does its best work: it splits availability into breakdowns and setup, and performance into small stops and slow running, so each drain maps to a specific countermeasure. A tall performance drain with a short availability drain points at speed and micro-stops, not breakdowns, a different fix entirely. A tall schedule-loss block above the OEE line is not an efficiency problem at all; it is a demand or staffing question, and attacking OEE will not touch it. The waterfall keeps you from spending improvement effort on the wrong bucket.

The waterfall doubles as an assignment map: each drain has a different owner, so the chart routes work instead of just scoring it.

What does the data say about the hours you can recover?

The recoverable hours are larger than most plants assume, because so few run near capacity. The Federal Reserve's G.17 Industrial Production and Capacity Utilization release put total U.S. manufacturing capacity utilization at about 75.8% in April 2026 roughly 2.4 points below its 1972–2025 long-run average. That figure is economic, not a plant OEE, but it points the same direction the waterfall does: a meaningful share of installed capacity is not turning into product. Inside a single asset, the schedule-loss block usually dwarfs every OEE loss combined, which is why the waterfall's top step often matters more to capacity than everything below it. For target-setting, calibrate against what a good OEE score means rather than a universal benchmark, and price the blocks in throughput terms so the chart drives a decision.

What makes a waterfall trustworthy?

A waterfall is only as honest as its smallest inputs. Two failures account for most misleading charts. The first is undercounted short stops: a supervisor remembers the 40-minute breakdown and forgets the twelve 90-second jams, so availability looks better than it was and performance loss absorbs the difference in the wrong bucket. The second is a drifting schedule boundary, when "planned time" quietly expands to hide downtime, OEE floats up without anything improving. Both are solved the same way: capture stops as they happen and fix the time model once, then never let either drift without a documented reason. A waterfall you cannot trust is worse than no waterfall, because it launders a fiction into a chart people believe. Harmony builds the waterfall from machine signals and operator-tagged reasons rather than end-of-shift recollection (see the platform), which keeps the small drains from disappearing into the big ones. A downtime-tracking discipline like the downtime tracking template is the manual version of the same idea. For proof the loss-accounting view holds up on a real floor, see the CLS field story and run your own numbers with the OEE calculator.