As the software improves, conjoint analysis is becoming an increasingly useful tool for analysing benefit segments among buyers. In particular, conjoint analysis is suited to industrial marketing research, because the small sample size arising from a small population of buyers is not a problem since analysis is fundamentally at the individual level. Moreover, the mathematical elegance of the technique promises that the findings will be impressively precise. Experience suggests, however, that a great deal of art, or informed judgement, is required to support the scientific application of conjoint analysis. Many pitfalls can catch the unwary researcher by surprise. Close collaboration between commissioning managers and researchers should minimise the pitfalls through an iterative learning process which allows the researcher to make decisions in the context of the particular market and allows the manager to understand the meaning and limitations of the findings. The paper discusses the lessons learned from three industrial marketing surveys, which are described at the outset. Decisions and problems encountered at the design, implementation and analysis stages of the surveys are considered. Practical suggestions for avoiding common flaws are given.