1) Confirm Model Design: It is not unusual for people to want to include as much detail as possible in their model. The reality is that you want to include only the detail necessary to achieve model’s objective.
To resolve this conflict in thinking we generate a Model Scope document. This document should include model scope, process map, relevant input data, assumptions, experiments to perform with model, etc. We need to have this document’s content agreed to by critical team members.
2) Fit Distributions to Data Collected: For model input (arrival rate, cycle times, etc.), we will pull data and/or collect data. Then we fit statistical distributions to this data. This ensures that we are correctly reflecting randomness in our model.
3) Select Runtime Parameters: To ensure that we are collecting output data from our model that accurately reflects steady state behavior of our system, we identify the model’s runtime parameters. These include length of warmup period, run length, and number of replications.
4) Verify/Validate Model: To ensure that we can trust the model to predict accurately, we will verify/validate that the model is behaving correctly. This includes performing sensitivity analysis on the critical input data.
5) Utilize Confidence Intervals: To ensure that our interpretation of the model results is correct, we utilize confidence intervals on critical output metrics.
