Cost structures and the movement of the innovation locus: a derived network approach

We consider the problem faced by a manager commissioning an innovative product requiring multi-stage sequenced innovation, when innovating agents have different costs and information transfer is expensive. We specify their optimisation problem and present a polynomial time solution method. We use the method to consider how cost networks influence centre choice switching by solving a series of stochastically generated networks and running logistic regressions on switching frequencies. The effect of expected innovation costs and its standard deviation are shown to be distribution dependent. Expected transfer costs are shown to have an unambiguous dampening effect on the amount of centre switching. Network size sensitivity is considered. Transfer costs are found to be far more influential on switching than innovation costs in a symmetric model. Cost trends that leave average costs unchanged are shown typically to have a significant non-zero effect on switching. A cost structure is introduced to model dichotomous expertise and to link innovation and transfer costs, and agent switches shown to be highly sensitive to an underlying learning cost measure. We then consider the set of sequences constrained to reach each possible final stage agent to reflect managerial specificity. Distributional parameters are found to have a dampened effect on within series changes, and their effect on cross series diversity is demonstrated to be opposite to that on within series changes.