The scalability of routing protocols for ad hoc networks is a key design issue for achieving large-scale deployments of MANETs. In this work, we propose S-Box, a ldquoblack-boxrdquo type scalability analysis framework. S-Box combines statistical design of experiments (DOE) methods, empirical modeling techniques, and a newly proposed performance measure, called the scalability index. The proposed scalability index is able to quantify the overall scalability performance of ad hoc routing protocols by aggregating the values of multiple individual metrics into a single scalar value. Given a specific deployment scenario, S-Box can be used to (1) determine the routing protocol that achieves the best overall scalability among a set of candidate protocols and (2) determine how to adjust parameter values (e.g., number of RREQ retries) of a routing protocol to achieve optimal scalability performance. To demonstrate its systematic operation and effectiveness, we use S-Box to study the scalability of DSR and AODV (with and without local repair), which shows S-Box can provide a convenient and systematic way to address practical scalability problems.