Frame the problem
A good model starts with a decision, not a diagram. Before you draw anything, pin down what your client is choosing, how they'll judge it and what the model has to contain.
This is the first of five pages that take one project from a client's question to the results you present. We use an urgent care clinic as the running example. Its numbers are made up for illustration.
Start from the decision
Clients rarely open with a decision. They say, "Our waits are too long." That's a problem. A model needs a choice to inform. Keep asking until you can fill in three lines:
| Item | Ask | Clinic example |
|---|---|---|
| The choice | What are you deciding between? | Add a third provider on weekdays, or add a fast-track room for minor cases? |
| What better looks like | How will you know it worked? | Most patients seen within 30 minutes, without adding more than one post |
| Who decides | Who makes the final call? | The clinic director, with the finance lead |
If nobody can name the choice, simulation is premature. Go back to your usual problem-solving tools, or see When to use simulation.
Pick the shape of the question
Most projects fall into one of a few shapes. Naming yours tells you what the model has to do well.
| Shape | The question | What the model needs most |
|---|---|---|
| Do we need more staff or space? | Whether to add people, rooms or equipment | Accurate capacity and demand at peak |
| Change a schedule or shift pattern | When people work, not how many | Real shifts and demand by time of day. Work schedules and arrival patterns need the Advanced view. |
| Compare two ways of working | Which of two processes works better | A careful check that the current process is right, so the comparison means something |
Choose 3 to 5 measures
Decide now how you'll judge "better". If you wait until you have results, you'll be tempted to pick the measure that flatters your favorite option. Give each measure a unit and, where the client has one, a goal.
| Your client says | Look for this in Quodsi |
|---|---|
| Cycle time, lead time, length of stay | Time in system on Overview, Flow time in Flow Analytics |
| Throughput, volume, patients seen | Throughput on Overview, Completed entities in Flow Analytics |
| WIP, backlog, people in the building | Avg WIP on Overview, Mean WIP in Flow Analytics |
| Bottleneck | Where's the constraint? in Flow Analytics |
| Cost, cost per case | Total cost on Overview, once you enable financial tracking on the Financial tab |
Some measures matter and can't be simulated, such as staff morale or patient satisfaction. Keep them on your list and say why the model doesn't cover them. Your client should hear that from you and not discover it later.
Work out what the model needs
For each measure, write down what the model must contain for that measure to mean anything. This one step keeps the model small and keeps you honest.
| Measure | The model must include | View |
|---|---|---|
| Door-to-provider time | Arrivals, triage, the queue for a provider | Basic |
| Provider utilization | Providers as a resource shared across steps | Intermediate |
| Peak-hour waits | Arrivals that vary by hour | Advanced |
| Cost per visit | Costs on activities | Basic |
If you can't connect a part of the process to a measure, it probably doesn't belong in the model. Some needs are judgement calls. A single provider count might do for a first pass, even though the real clinic runs shifts. Mark those as unclear and ask your client, so the gap doesn't get forgotten.
Draw the boundary
Decide where the model starts and ends, and write down what you are leaving out. Model the process, not the organization. In the clinic, you might start at the front door and stop at discharge, and leave out billing because it doesn't affect waits.
Start smaller than feels right. You'll learn more from a simple model that runs today than from a detailed one that runs next month. You can add detail where the results show it matters.
What you should have now
Before you move on, you should be able to put these on one page and have your client agree to them:
- The choice, what better looks like and who decides
- The first question a study could settle
- 3 to 5 measures, with units and goals
- What the model needs for each measure
- The boundary, and what is out of scope