A measurement is a structured observation that assigns a value, category or position to a property under a defined method.
Length, temperature, exam score, blood pressure, delivery time, customer satisfaction, population size and model accuracy are all measurements, but they differ in quantity, unit, scale, precision, method, uncertainty and comparability.
Quick answer: how should measurements be categorised?
- Quantity: what property is being measured?
- Unit: metres, seconds, dollars, percentage, score?
- Scale: nominal, ordinal, interval, ratio?
- Precision: how finely is the value resolved?
- Accuracy: how close is it to the relevant reference?
- Uncertainty: what range or error remains?
- Method: how was the measurement produced?
- Instrument: what device, test or rubric was used?
- Frequency: one-off, periodic, continuous?
- Provenance: who measured it, when, where and under which standard?
This article applies the wider framework from How to Categorise Anything to measurements as quantified or structured observations.
1. Separate measurement from the thing measured
A student is not a score; a machine is not a temperature reading. Measurements describe properties of entities or systems.
2. Separate quantity from unit
Length is a quantity; metres and centimetres are units.
3. Separate observation from interpretation
“38.5°C” is a measurement; “fever” is an interpretation under a threshold and context.
4. Nominal scales classify without order
Category labels such as blood type or document type may be measurements in a broad data sense even without numerical distance.
5. Ordinal scales preserve order
Low, medium, high or rank positions convey sequence but not equal intervals.
6. Interval scales preserve differences
Equal intervals are meaningful, while zero may not represent absolute absence.
7. Ratio scales include a meaningful zero
Mass, duration and count support ratio comparisons under appropriate units.
8. Scale type controls valid arithmetic
Not every numerical-looking code supports averaging or ratio interpretation.
9. Precision describes resolution
A value reported to three decimal places appears more precise than one reported to the nearest whole number.
10. Precision is not accuracy
A measurement can be highly precise and consistently wrong.
11. Accuracy depends on a reference
Reference standards, known values or validated methods provide the basis for accuracy claims.
12. Repeatability measures consistency
Repeated measurements under similar conditions should agree within expected variation.
13. Reproducibility extends across conditions
Different operators, instruments or laboratories may test whether the measurement remains stable.
14. Random error creates scatter
Repeated observations vary unpredictably around a central tendency.
15. Systematic error creates bias
A miscalibrated instrument can shift every measurement in one direction.
16. Uncertainty should be reported
Ranges, confidence intervals, tolerances and error bounds help prevent false precision.
17. Missing uncertainty is still uncertainty
An exact-looking number does not become certain merely because the error was omitted.
18. Direct measurements observe the property itself
A ruler measuring length is more direct than inferring length from an image under unknown scale.
19. Indirect measurements use proxies or models
Distance from travel time, body composition from impedance or economic activity from indicators depend on inferential relationships.
20. Proxy measurements need validation
The proxy must be shown to track the target property under the intended conditions.
21. Instrument identity matters
Device model, calibration status and software version can affect results.
22. Human-scored measurements need rubrics
Essay scores, observations and ratings become more comparable when criteria and assessor agreement are explicit.
23. Automated measurements need algorithm provenance
A model-generated score should identify model, version and relevant calibration.
24. One-off measurements capture snapshots
They may miss variability that appears over time.
25. Repeated measurements reveal trends
Longitudinal series can distinguish stable difference from temporary fluctuation.
26. Continuous measurement increases temporal resolution
Sensors can capture dynamics hidden by periodic sampling.
27. Sampling frequency should match system speed
Slow sampling can miss rapid changes or produce misleading summaries.
28. Aggregation changes meaning
Daily averages, monthly totals and peak values describe different aspects of the same underlying data.
29. Measurement granularity should match use
More decimal places are not useful when decision thresholds are broad or measurement uncertainty is large.
30. Units need controlled standards
Unit names, conversions and dimensional definitions should be explicit for interoperability.
31. Comparable measurements need comparable methods
Two values can use the same unit yet be incomparable because sampling, instruments or populations differ.
32. Normalisation changes representation
Per-capita, indexed and standardised measures can support comparison but should preserve the transformation method.
33. Threshold categories should preserve raw values
High, medium and low labels are useful operationally, but retaining the underlying measurement prevents information loss.
34. Measurement context matters
Ambient conditions, task difficulty, population and timing can alter interpretation.
35. Measurement drift requires monitoring
Instruments, rubrics and data pipelines can change gradually, shifting results without real-world change.
36. Version measurement definitions
Historical comparisons require knowing which formula, instrument, rubric or population definition was used.
37. A practical measurement record
- measurement ID;
- entity or system measured;
- quantity;
- value;
- unit;
- scale type;
- method;
- instrument;
- precision;
- accuracy evidence;
- uncertainty;
- time;
- location;
- operator or source;
- standard;
- version.
38. Measurement categories should improve comparability
A useful scheme helps determine whether two numbers can meaningfully be compared, combined or used in the same decision.
39. Measurement is disciplined representation
Every measurement compresses reality into a value under a chosen method.
40. The deeper idea
A number becomes meaningful only when quantity, unit, method and uncertainty travel with it.
To categorise a measurement well is to preserve what was measured, how, in what units, with what precision, and how much uncertainty remains.
Final answer
Categorise measurements by quantity, unit, scale, precision, accuracy, uncertainty, method, instrument, frequency, provenance and comparability. Keep raw observations separate from interpretations, report uncertainty, and version methods so changes in measurement do not masquerade as changes in reality.