This document discusses considerations for empirically analyzing algorithms through testing on real-world data and hardware. It notes that empirical analysis requires controlling many more factors than theoretical analysis, including the test platform, performance measures, benchmark data sets, algorithm parameters, and implementation details. It also discusses appropriate measures of performance like running time and operation counts, choosing representative test data sets, accounting for stochasticity in results, and using performance profiles to visually compare algorithms across data sets. The goal of empirical analysis is to provide insights into real-world algorithm performance that can complement theoretical analysis.