5.18 A good evaluation relies on good quality data. The evaluation questions will determine what data need to be collected, and when. This may be new data but will often also include monitoring data, that is, information collected and used as part of the ongoing policy delivery, describing the principal policy inputs and outputs (e.g. training sessions provided and completed). (For more information on planning and collecting monitoring data, see Chapter 7.)
5.19 Data requirements may also include data collected specifically for the evaluation through specially commissioned surveys and interviews with participants and frontline workers, and covering the details of the way the policy has been implemented. Evaluations of large scale policies might well also use data which already exist or are being collected for other purposes, for instance relating to local and regional economic conditions and performance (e.g. sectoral unemployment rates).
5.20 The specific data required for an evaluation will relate to the inputs, outputs, outcomes and impacts of the policy, and when these are expected to manifest. These will have been identified in the first step of planning the evaluation. Data collection processes will reflect the nature of the outcomes in question - outcomes which are unusual (e.g. impacts on individual economic wellbeing) or very specific to the intervention are likely to require special measurement through, for instance, dedicated surveys. Evaluation data may also relate to information about how the various elements of the policy are linked together, the actual delivery process and timescales.
5.21 Data collection will often need to commence before the policy is actually implemented, in order to ensure that the situation before the policy can be captured (also known as the "baseline"). Planning for data collection will obviously need to take place before this and so should be considered as early as possible. The timing of the data collection also needs to be considered carefully - eventual impacts of a policy may take many years to materialise, which are likely to be too distant to be collected as part of an evaluation project. In such cases it may be important to build in collection of data related to intermediate or proxy outcomes which can be used to measure impact in a shorter timeframe. These outcomes might then be "translated" into final outcome measures using the logic model framework.