Purpose: We answer the following questions under the constraints of our data set: Do brokers provide valuable information to their clients? Do they exhibit cross-sectional variations in trade execution ability? Is brokerage commensurate with ability (if any)? Originality: There is some research on execution ability between discount retail, full-service retail and institutional brokers, but not within the institutional space itself. There is also some work on the accuracy of broker analyst forecasts. However, to the best of our knowledge, there is no research that specifically measures the true value of brokerage services from actual institutional trading. Key literature / theoretical perspective: If brokers cross-sectionally exhibit varying levels of skill, then we should be able to observe this in a statistically significant way. By discretising the data, we can perform powerful statistical tests that are otherwise not possible, to discern whether patterns of broker skill exist in noisy trading data. Design/methodology/approach: We develop a new approach to discretising continuous trades and holdings data. Using this data representation, we investigate whether characteristics-based alphas on trades by specific brokers outperform other trades after controlling for manager skill. Similarly, we test whether particular managers exhibit superior execution performance relative to their peers. Finally, we determine whether a relationship exists between broker compensation (brokerage commission) and skill. Research limitations/implications: The sample size is small relative to the universe, though highly granular. Our results may not be representative of the industry as a whole. Practical and social implications: Brokers play a large role in the operation of fund managers. They provide both research and trade execution services, and incurs a small but significant cost to running the fund. This research elucidates the true utility of stockbrokers and potentially influences the way their services are priced.