This is a reference of manufacturing KPI formulas with one worked number each. The core one is OEE = availability x performance x quality, or good units x ideal cycle time / planned time. In the example shift below, that gives 72.0%. Before comparing any two figures, check that they use the same time base.
The example shift
Most formulas below use one hypothetical shift, so that you can trace each number. Where a KPI needs a longer period, the row gives its own numbers.
Need to know which KPIs to choose, and not how to work them out? Go to the list of production KPIs, which also shows how the losses of a shift add up.
OEE and time-based effectiveness
| KPI | Formula | Worked number |
|---|---|---|
| Availability | run time / planned time | 408 / 480 = 85.0% |
| Performance | total units x ideal cycle / run time | 18,000 x 1.2 s / 24,480 s = 88.2% |
| Quality | good units / total units | 17,280 / 18,000 = 96.0% |
| OEE | availability x performance x quality | 0.85 x 0.882 x 0.96 = 72.0% |
| OEE, short form | good units x ideal cycle / planned time | 17,280 x 1.2 s / 28,800 s = 72.0% |
| TEEP | OEE x planned time / all calendar time | 72.0% x 480 / 1,440 = 24.0% |
| OOE | good units x ideal cycle / staffed time | 345.6 min / 540 min = 64.0% |
The two OEE rows must give the same answer, which makes them a handy check on your data. Try it in the OEE calculator.
The last three rows look like rival scores, but they share one numerator. Only the time you divide by changes.
Output, flow and delivery
| KPI | Formula | Worked number |
|---|---|---|
| Throughput | good units / scheduled hours | 17,280 / 8 = 2,160 per hour |
| Actual cycle time | run time / total units | 24,480 s / 18,000 = 1.36 s |
| Takt time | available time / customer demand | 28,800 s / 18,000 = 1.6 s |
| Plan attainment | good units / planned units | 17,280 / 18,500 = 93.4% |
| Schedule adherence | jobs run as scheduled / jobs scheduled | 9 / 10 = 90.0% |
| On-time delivery | orders on time / orders shipped | 188 / 200 = 94.0% |
| Capacity utilization | actual output / maximum output | 86,400 / 108,000 = 80.0% |
| Lead time (Little's law) | work in process / throughput | 4,320 / 2,160 per hour = 2 hours |
| Value-added ratio | value-added time / lead time | 6 min / 120 min = 5.0% |
Cap plan attainment per product, so extra output on one item cannot hide a miss on another.
Takt time comes from demand, so it says what the line must do. Here the shift made one good unit every 28,800 / 17,280 = 1.67 s, slower than the 1.6 s takt.
Quality
| KPI | Formula | Worked number |
|---|---|---|
| First pass yield | good first time / units entering | 17,280 / 18,000 = 96.0% |
| Final yield | (good first time + reworked good) / units entering | 17,550 / 18,000 = 97.5% |
| Rolled throughput yield | FPY1 x FPY2 x FPY3 | 0.98 x 0.96 x 0.99 = 93.1% |
| Scrap rate | scrap units / total units | 450 / 18,000 = 2.5% |
| Rework rate | reworked units / total units | 270 / 18,000 = 1.5% |
| Defective PPM | defective units / total units x 1,000,000 | 720 / 18,000 x 1,000,000 = 40,000 |
| DPMO | defects / (units x opportunities) x 1,000,000 | 900 / (18,000 x 5) x 1,000,000 = 10,000 |
| Customer return rate | units returned / units shipped | 24 / 6,000 = 0.4% |
Scrap and rework rates share a base here, so 2.5% + 1.5% + 96.0% first pass yield = 100%. The guide to first pass yield goes deeper.
Downtime and reliability
| KPI | Formula | Worked number |
|---|---|---|
| Downtime % | stop time / planned time | 72 / 480 = 15.0% |
| Unplanned downtime % | unplanned stop time / planned time | 42 / 480 = 8.8% |
| Changeover share | changeover time / planned time | 30 / 480 = 6.3% |
| Mean time between stops | run time / number of stops | 408 / 24 = 17 min |
| MTBF | operating time / failures | 160 h / 8 = 20 h |
| MTTR | total restoration time / failures | 12 h / 8 = 1.5 h |
| Inherent availability | MTBF / (MTBF + MTTR) | 20 / 21.5 = 93.0% |
| Failure rate | failures / operating time | 8 / 160 h = 0.05 per hour |
| MTTF | total operating time / failed items | 48,000 h / 6 = 8,000 h |
| PM compliance | PMs done on time / PMs due | 46 / 50 = 92.0% |
The IEC vocabulary defines MTBF on operating time and MTTR as mean time to restoration, which includes waiting for parts and people. Others read it as hands-on repair only, so state which one you mean.
Inherent availability is not OEE availability, which also counts changeovers and other stops. Strictly, it uses hands-on repair time, so with waiting included in MTTR the result is closer to operational availability.
Work out your own with the MTBF and MTTR calculator.
People, cost, stock, energy and safety
| KPI | Formula | Worked number |
|---|---|---|
| Units per labor hour | good units / labor hours | 17,280 / 48 = 360 |
| Overtime share | overtime hours / regular hours | 96 / 1,600 = 6.0% |
| Absence rate | hours absent / hours scheduled | 40 / 1,600 = 2.5% |
| Cost per unit | (material + labor + overhead) / good units | 8,640 / 17,280 = 0.50 |
| Cost of poor quality | scrap cost + rework cost + other failure cost | 450 x 0.50 + 270 x 0.20 = 279 |
| Inventory turns | cost of goods sold / average inventory | 1,200,000 / 200,000 = 6.0 |
| Days of inventory | 365 / inventory turns | 365 / 6 = 60.8 days |
| Energy per unit | energy used / good units | 1,728 kWh / 17,280 = 0.10 kWh |
| TRIR | recordable injuries and illnesses x 200,000 / hours worked | 2 x 200,000 / 160,000 = 2.5 |
| LTIFR | lost-time injuries x 1,000,000 / hours worked | 1 x 1,000,000 / 160,000 = 6.25 |
Cost per unit falls with higher OEE only if the extra output gets sold, as explained in cost per unit. Compare a safety rate only with one on the same base, since 200,000 and 1,000,000 hours give very different numbers.
Five errors that break these formulas
- Averaging percentages. To combine lines, add up the minutes, then divide. See how to average OEE across machines.
- Mixing time bases. One report uses planned time, another uses run time, and the two can't be compared.
- Using top speed for some products only. Set an ideal rate for each product, or performance will drift when the mix changes.
- Counting rework as good. It pushes up quality and hides the time that repairs cost.
- Mismatched units. Cycle time in seconds and run time in minutes is a common source of a wrong answer, so convert before you divide.
How to use this sheet
Write the definitions you adopt into one page, including how breaks, changeovers and rework count. A formula only helps when everyone runs it the same way.
Then connect each number to a goal with a KPI tree. It will show which of these you really need.
Most of the inputs above are the same few numbers: time, stops and counts. Fabrico, a light MES for OEE, takes them live from PLCs or IoT sensors, with add-on AI cameras counting output where a machine has no usable signal, and shows OEE per line and shift. See how Fabrico calculates OEE.
What people ask next
What is the formula for OEE?
OEE = availability x performance x quality. It equals good units x ideal cycle time / planned production time, which is a useful cross-check. See availability, performance and quality for each factor.
How do you calculate MTBF and MTTR?
MTBF = operating time / number of failures. MTTR = total restoration time / number of failures. Use the same equipment and period for both, and try the MTBF and MTTR calculator.
What is the difference between first pass yield and quality in OEE?
Often none, when both count only units right the first time. They can differ if FPY is measured at a step while OEE quality is measured at the line's exit. See how to calculate first pass yield.
How do you calculate throughput?
Throughput = good units / time. Use scheduled hours for the line's real pace, or run hours to see its speed while running, and state which one you chose. See how to calculate throughput.
Can I use these formulas in Excel?
Yes. Put the raw inputs in their own cells, then let every KPI refer to those cells, so that one correction updates the whole sheet. The guide to the OEE formula in Excel shows the layout.
Fabrico is a light MES for OEE. It takes stops, speed and counts from PLCs or IoT sensors, with add-on AI cameras counting output where a machine has no usable signal, and operators tag the cause at the line.