28. There are several general conclusions which can be drawn from our work; none of them can pretend to any originality but they are worth mentioning as they deal with questions likely to arise in any large scale cryptographic job.
29. The first deals with refinements of technique. It was our experience (1) that any refinement in technique was worth while provided that it could be absorbed in the main system in such a way as to give little or no extra work (2) that by having a sufficiently comprehensive series of charts refinements usually could be so absorbed (3) that the total effect of a large number of minor improvements was very great. A good example of this was the evaluation of tetragram repeats between messages. Originally we took all tetragrams as equally valuable despite the fact that certain types of message were far more likely than others to give rise to such repeats owing to their containing a very high proportion of numerals. Later we divided the messages into 20 categories and within each category gave a separate evaluation according to whether the repeat began in the 1st, 2nd, 3rd ..... 10th, 11 - 30th, middle, last 30 letters of the text (i.e. 13 subdivisions per category). Once the scoring table had been compiled its use was a matter of seconds and caused no difficulty at all; the troublesome job was compiling and keeping up to date the table and it is this sort of thing that one is apt to shirk, but it is very well worth while.
30. The second point concerns the grade of labour that can be used on a given job. We found that by splitting up a job into as many separate parts as possible and by having a really adequate set of scoring tables quite elaborate computations could be brought well within the scope of any moderately intelligent clerical worker. The example in Para. 26 illustrates this; in the early days of Banburismus the cryptographers would have regarded such a calculation as too much trouble to be worth while - nevertheless in later days two Big Room girls would