3.6 Measurement function

What this heading covers

If the QME is the raw material, the measurement function is the recipe. The draft keeps the definition short: an algorithm or calculation performed on two or more quality measure elements. That brevity hides the decisions. A ratio needs a numerator and a denominator with the same population. A percentage needs a denominator that can reach the whole. A statistical function needs a distribution and a rule for outliers. The definition also implies that the function is fixed before the result is seen. A function chosen after the data arrives is a narrative, not a measure. The draft's Annex C names five measurement data types that a function can produce: count, time, ratio, percentage and statistical.

Record position: measurement concept 3 of 3, from the fifteen terms in clause 3.

What the heading asks of a team

What to put in place before publication

1. Write each function as a formula a reader can recompute, with the population stated in words.
2. Record the data type of the result, because a percentage and a ratio are read differently in a report.
3. Version every change to a function and keep the old results under the old version.
4. Test the function on a small fixture with a known answer before it goes into a dashboard.
5. State the rounding rule, since a percentage that rounds to one hundred is not the same as a clean run.

Where teams usually fall short

• A ratio computed from two registers with different cut-off dates, so the result moves with the extraction time.
• A percentage reported without its denominator, which hides that the population was four items.

What happens next

Annex C sets out the five measurement data types. Read it with this page, because the type decides how a result can be read.