4.1.2 Data Processors

Data processing takes a long time, and the time it takes varies enormously according to the complexity of the interviews: more complex interviews take longer to process. Ten pages per hour, which is usually about one and a half - two interviews per hour, sixty interviews per week, is an extreme maximum. If the interviews turn out to be much longer, then one must adjust the number of interviews per hour downward accordingly. The work is tedious and tires people quickly, so it is important not to overburden people working in this step. 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. If the interviews are long then the interviewers do fewer of them in a week. If the controlled vocabularies are extensive and complicated, data processing will be much slower than if the controlled vocabularies are simple and direct. For example, it is easier to code the violation acts from a narrative if there are only 6 kinds of violence in the controlled vocabulary than if there are 100 types. If there are more than a few types of violence, the data processors must constantly refer to their lists of violation types, which slows them to a crawl. The optimal classification of violations can be kept completely in the data processors' memories while they work on a given data source.

The ratio of interviewer time to data processor time can therefore vary considerably. Returning to the example from the previous example, how long will it take ten data processors to work through 5,000 interviews? If the violation controlled vocabulary is simple, ten data processors will be able to process about 600 interviews (sixty interviews per week per data processor X ten data processors) per week at a maximum. In their training period (which may last four-six weeks), the data processors may do half that number, or fewer. At that rate, they should be able to process all the interviews, with minor delays, in about nine weeks.

It is worth noting that the data processors can begin work as soon as there are sufficient interviews completed. To continue with the above example, the data processors might begin their work around the second or third week of interviewing. At that point many interviews should be complete, and the data processors should have sufficient work to occupy themselves. If they begin earlier, there is some risk that they will process interviews faster than the interviewers do them, which would leave them without anything to do. However, it is also important that the data processors not lag too far behind the interviewers because the data processors often identify inconsistencies in interviewers' work.