To monitor cellular networks, operators collect metrics such as throughput (TP) at the cell-level, but these are not a direct indication of user experience (UX). Since per-user metrics are unwieldy, we describe a new method to quantify UX from cell metrics. We propose a new cell metric, $N_{eff}$, that measures the effective number of competing users. We show how UX (burst rate) is obtained from TP and $N_{eff}$. This is used in the analysis of large field trial data from commercial networks.

Analyzing different test scenarios is difficult in live networks because of traffic inconsistency. We mitigate this with methods to extract consistent statistics, and jointly characterize changes in UX and TP across the entire network. We use these methods to quantify gains of a load balancing algorithm.