Data quality, sampling and collection

Reported figures are only as good as the chain that produced them, and that chain is usually longer than it looks: a field register, a tally sheet, a monthly aggregation, a database entry, a report. Each link can lose, duplicate or silently alter a number.

This section covers both halves of the problem. The first is quality — the dimensions along which data can fail, and the assessment routines that find failures before a funder does. The second is collection — how a sample is drawn and what it licenses you to say, how a question is worded and what that wording does to the answer, and how qualitative and quantitative evidence are combined into a claim stronger than either alone.