Skip to content
Data quality

Untrustworthy data must look untrustworthy.

A value is easier to evaluate when its condition, last update and reason are visible with it.

Stale

Stale

Shape
Dial with stopped hands
Shown as
Shown as “Last known value”, with its original sample time and its age.
01Failure modes

Three ways a number misleads.

Each one looks like an ordinary number until its condition is shown.

  1. In range, but out of date.

    A reading can look plausible and still be old. The value is kept, labelled “Last known value”, and shows its original sample time and its age.

  2. Present, but wrong.

    A reading that arrives can still be outside what is expected. It is shown with the Bad label and a specific reason.

  3. Shown after the connection failed.

    When a source connection is lost the last value can stay on screen. The loss is stated outright and the retained value is labelled.

02Anatomy of a reading

What a reading carries with it.

Healthy readings stay calm and readable. Exceptions become noticeable where they occur.

Process temperature

Machine 01 · AI_014
Stale
Last known value
185.0°C
No update since the last sample.
Last sample
10:42:18
Freshness
Age 2 m 40 s
  1. 01
    Name and source tagWhich measurement this is, and where it came from.
  2. 02
    Value and unitThe number, with its unit kept beside it.
  3. 03
    Condition label and iconBoth, always. Colour never carries the meaning alone.
  4. 04
    ReasonThe specific reason, when a reading is flagged.
  5. 05
    Last sampleThe original sample time. It never changes to imply a fresh reading.
  6. 06
    FreshnessHow old the last sample is.

Illustrative data.

03Evaluation

Five questions to ask of any reading.

Tick the ones your current tooling can answer. Useful for any platform you evaluate, including this one.

0 of 5 answered