For buyers contracting authority
Pick 72314000 when the contract's main job is to gather data and pull it together into one dataset, rather than to build software, run statistical analysis, or simply key in records that already exist on paper.
The boundaries that trip contracting authorities are all next door. Data capture (72313000) is the act of getting data off a source medium, often automated reading or digitisation, while collection-and-collation adds the step of consolidating disparate sources into a coherent whole. Data entry (72312000) is the manual keying of known records. Data analysis (72316000) is where interpretation and modelling live; reserve that code for contracts whose accepted output is findings, not an assembled dataset.
Where a single contract spans gathering, cleaning and analysing, set the primary CPV code by the dominant deliverable. If the authority is paying chiefly for a consolidated, validated dataset it can hand to its own analysts, 72314000 is the right pick. If the accepted output is the analytical conclusion, look to 72316000 instead.