Industry publishes relatively few scientific papers. Consequently, it is usually believed that bibliometrics is not very well suited to measure industrial science. The present article tries to assess the usefulness of bibliometrics for measuring industrial scientific activities. 11 814 papers and 84 658 patents originating from 199 multinationals are statistically analyzed in order to understand (1) the importance of industrial publications, (2) the fields of science privileged, (3) the level of science useful to industry, and (4) the science-technology relationships.