Five level of measurement mistakes in research reports

Five mistakes with levels of measurement that turn up in real research reports, each with the fix. They cover nominal codes treated as numbers, ratio claims on rating scales and on Net Promoter Score, ordinal bands charted without their order, and multiple-choice shares added together. Use it as a last check on a survey report before it goes to stakeholders. Koji picks summaries that match each question type, which avoids most of these by default.

Where this appears

• Levels of Measurement: Which Statistics Each Survey Question Type Allows (2026)

What's in this guide

1. Averaging category codes: If regions are coded 1 to 4, an average region of 2.3 describes the coding scheme. Nominal codes support counts only.
2. Ratios on rating scales: Twice as satisfied depends on where the scale starts. Report the gap in points, or in the share at the top of the scale.
3. Percent change on an arbitrary zero: NPS going from 10 to 20 is a 10-point gain. Calling it doubled is a ratio claim on a scale with no true zero.
4. Dropping the order in banded answers: Size, tenure and frequency bands are ordinal. Show the cumulative share at or below each band.
5. Adding up multiple-choice shares: Each option is its own yes or no, so shares often total more than 100. Don't add them or put them in a pie.