This dissertation studies firms' strategic interactions in anticipation of random service disruption following technology failure. In particular it is aimed at understanding how contracting decisions between a vendor and one or multiple clients affect the firms' subsequent decisions to ensure disruption response and recovery are managed as efficiently as possible. This dissertation consists of three studies that were written as standalone papers seeking to contribute to the literature on contract design and technology management in operations management. Together, the three studies justify the importance of structuring the right incentives to mitigate disruption risks.

In the first study we contribute to this literature by means of an analytical model which we use to examine how a client and vendor should balance investments in response capacity when both parties' efforts are critical in resolving disruption and each may have different risk preferences. We study the difference in the client's optimal expected utility between a case in which investment in response capacity is observable and a case in which it is not and refer to the difference in outcomes between the two cases as the cost of complexity. Firstly, we show that the cost of complexity to the client is decreasing in the risk aversion of vendor but increasing in her own risk aversion. Secondly, we find that a larger difference in risk aversion between a client and vendor leads to underinvestment in system uptime in case the client's investment is observable, yet the opposite happens when the client’s investment is not observable.

In the second study we further examine the context of the first study through a controlled experiment. We examine how differences in risk aversion and access to information on a contracting partner’s risk preferences interact in affecting contracting and investment decisions between the client and vendor. Comparing subject decisions with the conditionally optimal benchmarks we arrive at two observations that highlight possible heuristic decision biases. Firstly, subjects tend to set and hold on to an inefficiently high investment level even though it is theoretically optimal to adjust decisions under changing differences in risk preferences. Secondly, subjects tend to set and hold on to a penalty that is too high when interacting with more risk averse vendors and too low in case the vendor is equally risk averse. Furthermore, cognitive feedback on the vendor’s risk aversion appears to have counterproductive effects on subject’s performance in the experiment, suggesting cognitive overload can have a reinforcing effect on the heuristic decision biases observed.

In the third study we construct a new analytical model to examine the effect of contract design on a provider's response capacity allocation in a setting where multiple clients may be disrupted and available response capacity is limited. The results show that while clients may be incentivized to identify and report network disruptions, competition for scarce emergency resources and the required investment in understanding their own exposure may incentivize clients to deliberately miscommunicate with the vendor.