P3.3—Decisions about measurements and observations

The number and distribution of measurements should reflect the question, expected variation and available range. Repeat an anomalous result, but do not delete it silently.

Use a broad enough range to reveal a trend, replicate suspected anomalies, and select reagents whose observations distinguish between candidate ions or compounds.

If one titration is far from two concordant titres, repeat it rather than averaging it automatically. If two ions give the same first test, choose a second reagent that produces different observations.

“Repeat until the answer looks right” is not valid method design; the rule for accepting or investigating data must be stated.