For in-store designs one has to define a standard of measurement. In most cases it is no use to design something with the use of only one store. It is preferable to use a reasonable number of stores. Great attention should be paid to the unsolvable problem of comparability between stores. It is preferable to use the more simple research designs instead of the sophisticated ones, because if something goes wrong, and in most cases it does, one need not throw the full experiment away. One should try to design an experiment that takes exogenous factors into account like seasonal fluctuations, trends, etc. It pays to devote great attention to the human factor involved by motivating people who participate, but not too strongly. Rigid field control is an absolute necessity in order to reach honest evaluation.
- This could also be of interest