Sales Forecasting System

Transcribed from the printed issue; not yet reviewed. Read it in the PDF of the issue, page 60

Scope¶

The system was developed to allow production to be more closely geared to the changing market. The products were highly seasonal. Forecasters wanted to use their knowledge of hierarchical structure of the Nestle products (e.g. to improve estimates of frozen fish sales in light of projected trends in whole frozen food sector). To complicate the whole process, divisions had different accounting periods (some monthly, some 4-weekly).

History¶

Some of the talk outlined the evolution of the system through computers, APL’s and Operating Systems. Briefly, the original system (using VSAM under CMS) did not allow multi-user update, so they moved to SHAREFILE (a ‘crash-resistant’ filing system). They are considering moving to APL2 and DATABASE (DB2) in the future.

The original aim was to use the APL development as a prototype for a more formal COBOL system, as the need for an evolutionary approach was foreseen. One year after the system had been finalised, the transfer to COBOL was abandoned and the APL system installed in four weeks! Compatibility with other company systems was maintained by continuing to create VSAM files in the same format as the previous system, rather than altering all dependent systems.

Algorithms¶

The aids for producing forecasts appeared flexible, including options to: (i) view and alter forecasts on a 4-weekly or a monthly basis or by normalised months. (ii) use the system as a simple calculator (e.g. to reduce forecasts by 10% or add 10). (iii) alter any elements and row/column totals in a 2-dimensional display; the system would jiggle it about to find a sensible result.

Conclusion¶

This system seems to embody many of the features that we are attempting to put into our own sales forecasting system, so it can be done . . . .