Purpose: Due to the complexity of digital services, companies are increasingly forced to offer their services “in permanent beta”, requiring continuous fine-tuning and updating. Complexity makes it extremely difficult to predict when and where the next service disruption will occur. The authors examine what this means for performance measurement in digital service supply chains. Design/methodology/approach: The authors use a mixed-method research design that combines a longitudinal case study of a European digital TV service provider and a system dynamics simulation analysis of that service provider's digital service supply chain. Findings: With increased levels of complexity, traditional performance measurement methods, focused on detection of software bugs before release, become fragile or futile. The authors find that monitoring the performance of the service after release, with fast mitigation when service incidents are discovered, appears to be superior. This involves organizational change when traditional methods, like quality assurance, become less important. Research limitations/implications: The performance of digital services needs to be monitored by combining automated data collection about the status of the service with data interpretation using human expertise. Investing in human expertise is equally important as investing in automated processes. Originality/value: The authors draw on unique empirical data collected from a digital service provider's struggle with performance measurement of its service over a period of nine years. The authors use simulations to show the impact of complexity on staff allocation.

Resource pooling is known to benefit performance through reduced congestion, but primarily in settings with homogenous demand. In settings where demand is heterogeneous, pooling can be counter effective. The effects of pooling of staff when demand is heterogeneous and dependent are not known. We present a simulation model based on a service supply chain that delivers Interactive TV to customers. Customers expect high performance in terms of innovativeness and reliability. Based on the results of simulation analysis, we find that when target innovativeness of the service is increased, pooling outperforms not pooling, but the delays that are involved with pooling will make the system and hence its performance unstable. Stable and high performance can be realized through “unbalanced” hiring. This means that a target performance increase in the upstream stage of the chain (innovation), is accompanied by hiring staff in the downstream stages of the chain (QA and operation).