This descriptive study explores the reasons for integration of the personal selling function, i.e., the use of employee ("direct") salespeople rather than manufacturers' representatives ("reps"). A hypothesized model is developed based on both transaction cost analysis and the sales force management literature. Data from 13 electronic component manufacturers covering 159 U.S. sales districts are used to estimate a logistic model of the probability of "going direct" in a given district. Results are shown to be stable across specification and estimation methods and to fit the data well. The transaction cost model is generally supported. The principal finding is that the greater the difficulty of evaluating a salesperson's performance, the more likely the firm to substitute surveillance for commission as a control mechanism, i.e., to use a direct sales force. Among other findings, direct sales forces are also associated with complex, hard-to-learn product lines and with districts that demand considerable nonselling activities. Several factors prove unrelated, including company size, the amount of travel a district requires, and the importance of key accounts.