How to Improve OEE: The 5 Levers That Actually Move the Number

OEE is the most commonly cited manufacturing metric, and the most commonly misunderstood. The formula itself is simple: Availability x Performance x Quality. But the definitions of each component determine whether your OEE number is useful or misleading.

Availability = Run Time / Planned Production Time. Planned production time excludes scheduled breaks, planned maintenance, and periods when the line isn't scheduled to run. The key question is whether you're counting changeover time as an availability loss (you should) and whether you're counting minor stops under 5 minutes (many plants don't, which inflates the number).

Performance = (Total Parts / Run Time) / Ideal Cycle Time. This is where hidden losses live. If a machine is rated for 120 parts per minute but you're running it at 105 because "that's where it runs best," you're carrying a 12.5% performance loss that most plants never question. The rated speed should be the nameplate speed or the best demonstrated speed, not the current operating speed.

Quality = Good Parts / Total Parts. This includes startup scrap, in-process defects, and anything that doesn't meet spec on the first pass. Reworked parts that eventually pass should still count against quality because they consumed capacity without producing sellable output.

The multiplication effect is why OEE is so hard to improve. If each component is at 90%, OEE is 72.9%, not 90%. Small losses in each category compound. A 3% improvement in availability, a 2% improvement in performance, and a 1% improvement in quality might not sound like much individually, but combined they can move OEE by 5-6 percentage points, which translates to significant additional production capacity.

The first step to improving OEE is getting the measurement right. If your OEE is above 90%, either you have a world-class operation or your definitions are too generous. Most plants that claim 90%+ OEE are excluding losses that should be counted [2].

Ask any machine operator: are you running at rated speed? The honest answer in most plants is no. And the reasons are usually practical: running at full speed causes more jams, produces more scrap, or wears tooling faster. The operator found a speed that "works" and nobody questioned it.

The problem is that a 10% speed reduction across an 8-hour shift is 48 minutes of lost capacity. Multiply that across 5 lines and 250 production days, and the plant is giving up the equivalent of 25 production days per year without anyone tracking it as a loss.

Common sources of performance loss:

• Chronic rate reduction: the machine was slowed down 6 months ago because of a quality issue. The quality issue was fixed, but nobody restored the speed. This happens more often than anyone admits. Walk the floor and compare actual cycle times against the rated cycle times in your production system. Discrepancies of 5-15% are common.
• Speed variation within a run: the machine starts at rated speed, but the operator slows it down periodically to "let it catch up" or prevent a known intermittent issue. These micro-slowdowns don't register as stops but reduce average throughput by 3-8%.
• Idling and minor stops: the machine is powered on and in cycle, but waiting for material, waiting for downstream equipment to clear, or running empty because the infeed is starved. These are not availability losses because the machine isn't technically down. They're performance losses because the machine is cycling without producing output.
• Tooling wear: as cutting tools, forming dies, or sealing bars wear, cycle time increases to maintain quality. A fresh tool might run at 100% speed. The same tool at 80% of its life might be running at 92% speed. If tools are changed on a calendar rather than on condition, the tail end of the tool life drags down average performance.

The fix starts with transparency. Display the actual vs. rated speed on the HMI or production board for every machine. When the gap is visible, the conversation about why it exists happens naturally [3].

OEE data is typically viewed per-machine or per-line as a time-series chart. Machine 4 ran at 72% OEE yesterday. The trend over 30 days is flat. That's useful but incomplete.

What's missing is spatial context. Machine 4's OEE dropped because performance fell to 88% in the afternoon. The time-series chart shows the drop. It doesn't show that machines 3, 4, and 5 all experienced performance drops between 1 PM and 3 PM, that they're all in the south-facing bay of the building, and that ambient temperature in that bay exceeds 30 degrees C every afternoon in summer.

Spatial OEE analysis overlays availability, performance, and quality metrics onto the physical floor plan. This view answers questions that per-machine analysis cannot:

• Are availability losses clustered in one area? (Points to shared infrastructure: power, utilities, environmental conditions)
• Do performance losses correlate with physical proximity? (Points to upstream/downstream dependencies or environmental factors)
• Do quality losses on one machine correlate with the output variation of the preceding machine? (Points to process interaction effects)

The practical approach is straightforward. Export your OEE data with machine location coordinates. Color-code each machine on the floor plan: green for above target OEE, yellow for within 10% of target, red for below. Update daily. Within a week, spatial patterns become visible.

Plants that adopt spatial OEE views commonly find that 20-30% of their total OEE losses are driven by environmental or infrastructure factors that are invisible in per-machine analysis. Fixing an HVAC zone, rebalancing an electrical feeder, or addressing a shared compressed air pressure drop can improve OEE across multiple machines simultaneously.

The compounding effect is significant. If a single environmental fix improves performance on 5 machines by 3% each, the aggregate capacity gain is equivalent to improving one machine by 15% but much easier to achieve because you're addressing one root cause instead of five separate symptoms.