We were interested in understanding how people collaborate in the preparation of documents, so that we could make more informed predictions about the impact of new technologies on document preparation. We were particularly interested in learning if there were any recurring patterns in such collaborations, and in gaining a better understanding of time as a resource in document-centred work.
We conducted the study at "The Commission", our name for a large financial institution in Washington DC. A preliminary ethnographic study, based largely on interviews, provided us with a general understanding of the work. We then conducted a week-long diary study of 25 members of staff in four separate devisions. At the start of each day we obtained from each person a plan of the activities they intended to carry out that day, and at the end of each day we conducted an interview of 20 to 30 minutes during which we constructed a diary of the day's actual activities. The daily plan was useful in identifying activities that they might otherwise have forgotten to include in the diary, and in helping us separate planned from unanticipated tasks.
In our analysis, we traced the progress of a number of documents during the week of study. A large proportion of the 300-odd documents mentioned served only as sources for reference. However, we found about 20 examples of complex processes involving several members of staff in the preparation of one document. We traced the progress of these documents by constructing flow charts. A number of recurring patterns could be observed, such as the tendency of senior staff to delegate the drafting of short documents to junior staff, and then to spend a final period in rapid iteration between discussion (by both authors) and revision (by the junior staff member).
We also identified a phenomenon which we called last-minute authoring: individuals and teams appeared to be skilled in estimating how long a document's preparation would take, and would usually start the authoring process just in time to complete the document by the assigned deadline. Sometimes, however, technology failures would cause deadlines to be exceeded. Often an author would appear to use any excess time "tinkering" with the text, rather than attending to other work.