Within the CISM project empirical models serve as baselines against which to monitor the increase in predictive skill of physics-based numerical codes as they advance. Empirical models also establish a suite of forecast models from which to evolve toward greater forecast accuracy by incorporating numerical codes as they advance enough to increase forecast skill. Establishing a suite of forecast models allows the CISM project to contribute results of use to operational space weather forecasting earlier than might otherwise be possible. Developing a suite of empirical models allows CISM to address issues of data ingestion early in the project under relatively simple conditions. Out of the data-ingestion effort has come a technique of probabilistic forecasts, which allows one to ingest information on IMF Bz from solar data instead of from measurements taken only at the Lagrangian L1 point. To implement the evolution from empirical to numerical forecast models, CISM has adopted a coupled two-line system of model development, a science-model line interacting with a forecast-model line.