DataOps
Check data quality before calculating OEE
Availability, performance and quality must use consistent production boundaries.
How do you design a production dashboard? Which metrics should it show?
Short answer
Design a production dashboard for one decision and one audience. Show three to five metrics, each with its target and recent trend. State units, time range and data freshness clearly. If a value is missing, show it as missing, not as zero.
One decision, one audience
A dashboard should answer one question, such as 'Is the line on target at the end of this shift?' Name the reader first. A shift supervisor, a maintenance planner and a plant manager need different views, and one screen that tries to serve all of them usually serves none well.
A metric means little without its target and recent trend. Take a hypothetical value of 74 %: on its own it says little. With a 75 % target and a last-week trend, the reader can see whether the line is close to target or moving away from it. Show data freshness too, such as the last update time and any gaps, so readers know how far to trust the screen.
A hypothetical shift board
Hypothetical shift board for one line: four metrics, OEE with a 75 % target and a 14-shift trend, availability, total stop minutes and good-part ratio. Each chart's title carries its unit and time range, for example 'OEE, %, last 14 shifts'. Shifts with no data are left empty and labelled 'no data' rather than drawn as zero, because zero would say the line stopped. Two measures with different units go on separate charts, not on a dual axis. Forge can turn a CSV table of this kind into a single-file HTML dashboard, but the design choices remain yours.
Typical dashboard mistakes
The most common mistake is putting everything on one screen. A board with twenty charts helps nobody decide. The second is dual-axis charts: two units on two scales in one chart can make unrelated lines look connected. The third is decorative 3-D charts and rainbow colour scales, which make values hard to read and can hide differences. The last is showing a missing value as zero, which suggests the line stopped when the data simply never arrived.
Follow-up questions
How do I decide who the dashboard is for?
Write one sentence: who reads it and what they decide from it. If you cannot write that sentence, you probably have two dashboards. A management summary and an operator screen are usually separate pages, even when they draw on the same data.
How fresh should the data be?
That depends on the decision the dashboard supports. A control screen needs data current enough to act on, while a weekly management report can use older data. Whatever the refresh rate, show the last update time so readers know what they are looking at.
Sources
Example numbers are hypothetical, not measured field results.
Related questions
All questionsHow do you clean sensor CSV data?What is a data historian, and how does it differ from a SQL database?What is a 4-20 mA signal, and why does it start at 4 mA, not 0?How is a 4-20 mA signal scaled to engineering units?How often should an instrument be calibrated?How can I detect sensor drift in an industrial process?