Academic inventions have to be transferred to industry to become an innovation. Scientists face multiple options for this transfer, from informal knowledge transfers to patents, licences, and spin-offs. These transfer channels require different efforts and inhibit different degrees of complexity. We want to theoretically explain the inventor's choice of a certain transfer channel. Under the assumption that (i) dealing with complexity is similar to facing risk, and (ii) scientists are risk averse, we show that the chosen transfer channels are path-dependent: with increasing commercialisation experience inventors choose more complex channels, up to a certain limit of complexity.