In spite of the fact that there exist newer data compression algorithms with better compression ratios, the Box Car Back Slope (BCBS) algorithm is still offered by vendors of data acquisition systems and remains in use in many process industries. The assumption behind the BCBS algorithm is that the recorded process variable can be approximated by a piece-wise linear function, whereby variable values within a threshold window around the line segments are not stored. The window size is a critical parameter which should be selected such that a desired compromise between the data compression ratio and the approximation error is achieved. Based on a criterion function reflecting such a compromise, an automatic, recursive algorithm for tuning the BCBS threshold window size is presented in this paper. The algorithm has been used successfully to tune the window size for hundreds of process variables at the carton board manufacturing plant of Assi-Doman Frovi, Sweden. (C) 2003 Elsevier Ltd. All rights reserved.