It has been observed that IEEE 802.11 commercial cards produced by different vendors show a different behavior in terms of perceived throughput or access delay. Performance differences are evident both when the cards contend alone to the channel, and when heterogeneous cards contend together. Since the performance disaligment does not disappear by averaging the environmental factors (such as propagation conditions, laptop models, traffic generators, etc), it is evident that the well known throughput-fairness property of the DCF protocol is not guaranteed in actual networks. In this paper we propose a methodological approach devised to experimentally characterize the IEEE 802.11 commercial cards thus understanding and predicting their performances in different network scenarios. We set up some specific experiments using a custom test equipment, able to classify the card behavior not only in terms of figures which are evident to the user perspective (such as the throughput), but also in terms of low-level channel access operations and delays. Our approach is able to detect potential hardware limits or not-standard MAC implementations, which severely affect the contending card performance.