PlantOps BI

OEE Across Lines

A common way to build OEE in DAX is to multiply Availability, Performance and Quality, each summed in units. On one line that is exact. Across lines with different ideal rates it drifts. On the demo dataset the company figure reads 65.1% when the lines actually ran at 62.8%.

The Numbers

Every figure below is computed from fact_production_hourly (155,340 line hours) when the site is built. Each hour is 60 planned minutes, and run time is 60 minus the downtime minutes.

By line. On a single line both methods give the same OEE, because there is only one ideal rate.

LineFamilyIdeal rate (uph)AvailabilityPerformanceQualityOEEShare of units
PDX-01 Enclosures 50 87.3% 73.4% 95.8% 61.3% 2.8%
PDX-02 Enclosures 43 86.7% 71.1% 95.8% 59.0% 1.4%
PDX-03 Panel Assemblies 28 86.7% 68.6% 94.8% 56.4% 0.9%
PDX-04 Brackets 175 86.6% 74.9% 97.7% 63.4% 6.0%
PDX-05 Cable Management 85 86.7% 81.4% 98.2% 69.3% 3.2%
BOI-01 Fasteners 693 87.5% 75.5% 98.8% 65.3% 39.9%
BOI-02 Fasteners 640 86.9% 82.4% 98.8% 70.8% 24.2%
BOI-03 Brackets 182 86.8% 61.3% 97.8% 52.0% 5.1%
BOI-04 Cable Management 87 87.1% 73.7% 98.2% 63.1% 3.0%
RNO-01 Enclosures 44 86.9% 80.2% 95.8% 66.8% 1.6%
RNO-02 Panel Assemblies 22 87.0% 72.1% 94.8% 59.5% 0.7%
RNO-03 Brackets 194 87.6% 76.0% 97.8% 65.0% 11.2%

Rolled up. Unit-pooled multiplies components summed in units; time-based sums every component in minutes at ideal rate.

LevelUnit-pooled QualityUnit-pooled OEETime-based QualityTime-based OEEOverstated by
Portland Plant 97.1% 63.3% 96.4% 61.8% 1.5 points
Boise Plant 98.7% 65.7% 98.5% 63.1% 2.5 points
Reno Plant 97.4% 64.8% 96.4% 64.0% 0.8 points
Company 98.3% 65.1% 97.1% 62.8% 2.3 points

Try a Different Mix

Pick the lines to roll up. The browser recombines their totals both ways, the same arithmetic as the tables above.

Lines in the roll-up

Unit-pooled OEE
65.1%
Time-based OEE
62.8%
Difference
2.3 points

Why It Drifts

OEE is a ratio of time: the minutes spent making good parts at ideal speed, over planned minutes. Summing units instead of minutes weights each line by how many parts it makes, not by how long it ran. The two fastener lines make 64% of all units in 19% of the planned time, and they also have the best quality. Unit-pooled Quality is mostly their quality, so it reads 98.3% against 97.1% measured in time.

The direction depends on the data. Pooling overstates here because the fastest lines also have the best quality; where the high-volume lines are the worst, it understates. Either way the figure stops meaning time.

Availability is already in minutes, so it is the same either way. Performance and Quality are where the weighting slips in, and the error compounds because the three are multiplied.

The correct company number is the planned-time-weighted average of the line OEEs. A plain average of the twelve lines gives 62.7%, which is closer but still wrong, since a two-shift line counts as much as a three-shift one.

DAX That Adds Up

Put every component in minutes at the line's ideal rate. Then Availability x Performance x Quality equals OEE at the line, the plant and the company, and the three components still explain where the time went.

The two ideal-minutes measures iterate the distinct ideal rates rather than the fact rows. CALCULATE turns each rate into a filter, so the storage engine sums units for every line at that rate at once, and DIVIDE runs 12 times instead of once per hour of history. Two lines that share a rate are summed together, which gives the same minutes as dividing each line separately.

Planned Minutes = COUNTROWS ( fact_production_hourly ) * 60

Run Minutes =
[Planned Minutes] - SUM ( fact_production_hourly[downtime_minutes] )

-- Minutes the units would take at the line's ideal rate, summed per rate
Ideal Minutes Produced =
SUMX (
    VALUES ( dim_line[ideal_rate_uph] ),
    DIVIDE (
        CALCULATE ( SUM ( fact_production_hourly[units_produced] ) ) * 60,
        dim_line[ideal_rate_uph]
    )
)

Ideal Minutes Good =
SUMX (
    VALUES ( dim_line[ideal_rate_uph] ),
    DIVIDE (
        CALCULATE (
            SUM ( fact_production_hourly[units_produced] )
                - SUM ( fact_production_hourly[scrap_units] )
        ) * 60,
        dim_line[ideal_rate_uph]
    )
)

Availability % = DIVIDE ( [Run Minutes], [Planned Minutes] )
Performance % = DIVIDE ( [Ideal Minutes Produced], [Run Minutes] )
Quality % = DIVIDE ( [Ideal Minutes Good], [Ideal Minutes Produced] )

-- Equals Availability x Performance x Quality at every level
OEE % = DIVIDE ( [Ideal Minutes Good], [Planned Minutes] )

The site's tests check that these components multiply to the time-based OEE for every line, plant and the company, and that the company figure equals the planned-time-weighted average of the lines. The figures move only if the demo data changes.

OEE definition and benchmarks · About the demo dataset