How much and what type of know-how information should an organization collect and remember? We consider this question and derive a number of conclusions. First, information about successes is more useful than information about failures, explaining companies' penchant for ``best practice''. Furthermore, the optimal amount of information is maximal for large organizations faced with medium levels of environmental stability and high variance of potential payoffs. We also derive a strong competency trap: remembering information about mediocre practices may reduce profits since it inhibits exploration. We show that firms with a high turnover per time period get more benefit from an information system, but, as a consequence of the competency trap, they should be more critical of codifying knowledge. Moreover, it may be optimal to communicate some information on a plant level but to not disseminate it on a firm level.