that are at the same level in the classificatory hierarchy. For example, instead of (two categories of) (a) torture, and (b) beating, a list should be created that has torture with various sub-categories, such as beating and electric shocks, etc.45 This is already a fairly well-established practice in some Zimbabwean organisations involved in human rights data collection. Other lists of controlled vocabulary would include amongst others, regions of the country, State security agencies and proxy forces and so on. It is important to remember that data processors apply the definitional decisions to the raw information that is available. This is the point at which there must be as much standardisation and consistency in terms of interpretation and definition – and the project management will have to assess the integrity of the coding process on an ongoing basis. This also underscores the importance of ensuring comprehensive training and ‘dummy runs’ before the processing is initiated. This is tiring and time-consuming work. If the controlled vocabulary is complicated, the data processing will be much slower. This will obviously impact on the time taken to complete the project. At this stage it is impossible to give an accurate prediction on how long it would take to process 7,000 statements or to sufficiently train a team of coders and processors. A very general estimate based on previous experiences would be that ten data processors would take between four to six months to complete the task. This would probably be at the slower end of the ‘market’. This involves coding, as well as entering the data into the database. In many projects, these functions are separated. It took eight data coders and five processors over 15 months to process 8000 statements in Timor Leste, yet elsewhere others have argued that at ‘full steam’ a data processor can process as many as 10 to 12 statements a day. “As a very general rule of thumb, one data processor can do as many interviews in a day as an interviewer can do in a week.”46 Database representation The database must be able to represent all the information in the processed data, and must be able to produce the analytical reports that you are seeking to engage. In other words, it will be necessary to draw up a list of questions that you think the available data should be able to address. Having the data already available, however, does allow for the database design and construction process to be more tailored around the existing information sources. In this regard, it would be necessary to take samples of individual cases (narratives and so on) to determine exactly what detail of information has been captured. Inevitably, large-scale databases often require further changes to the database design as additional fields of information are identified. This is often an ongoing process in situations where the data is being gathered from scratch. In this scenario, however, much of this can be avoided from the initial stages of the project. It is recommended that the project engage a professional human rights relational database design expert to facilitate the database design process.47 The project will need an in-house database programmer to handle the plethora of issues that inevitably arise. The project should consider how best the external professional can impart skills and expertise to the in-house programmer. 45 ibid., p. 33. ibid., p. 74. 47 A more detailed exposition of database design is available on request. 46 37