There are mutliple solutions, one of which is found stored in solution.data.
Jimmy Lee models each visit of a shopper by a unit square. The square $S_ij$ denotes the j-th visit of the i-th shopper. We then need to place 140 squares into a 35*4 table. That square $S_ij$ is placed at position (p,q) of the table means that the saleslady p is visited in the q-th week by the i-th shopper as his/her j-th visit. A diffn constraint is used to enforce that the squares do not overlap. We also impose some other constraints, such as element and alldifferent, to meet the rest of the requirements.
Using this modeling, the median execution time and the mean execution time with the best and worst runs removed for E-GENET are 0.663 second and 0.725 second respectively while CHIP does not terminate within 10 hours. H. Simonis later developed a CHIP program for solving the problem using another modeling with only the among constraint. This program is also able to solve the problem almost instantly. The efficiency of Simonis’s program is derived from a domain specific value ordering heuristics tailor-made for this problem.