Cleaning and analyzing data

Incident logs, survey exports, a spreadsheet three people have been adding to for two years with slightly different spellings of the same town. The cleaning takes longer than the analysis, and it is what gets skipped when the report is due.

AI is useful here. It spots the four spellings of one place name, suggests a structure, and writes the formula you could not remember. It is far worse at arithmetic than it looks: it will give you a total that is confidently wrong and never mark it as a guess.

Watch out for this

A spreadsheet leaves an organisation easily, because a file feels like data rather than like people. It is people: names, phone numbers, villages, case outcomes, in columns you were not thinking about. Save a copy, delete the identifying columns from it, work only from that, and check every number against the sheet.