Test your changes

With a Baseline you trust, you can test ideas in minutes. Choose a few changes, run them side by side, and let the numbers show which ones hold up.

Start with your Baseline

Every comparison is against the Baseline, the model as it stands today. If you haven't run it yet, open What-if Analysis and click Create Baseline scenario.

Before you change anything, read Flow Analytics. Under Where's the constraint?, the top row is the constraint. Then read two columns closely:

• Constrained by says what is limiting the step: its own capacity, a resource it needs, or both. If it names a resource, adding capacity to that step won't help. This one column can save you a wasted scenario.
• Reps agreeing says how many replications, or repeated runs, reached the same verdict. A higher number means a more reproducible finding.

Choose what to try

Pick 1 to 4 changes that could affect the choice your client is making. Say each in your client's words, such as "A second scanner". Include a cheap option as well as the expensive one, and include the option your client doesn't expect. Sometimes changing nothing is the right answer, and you can only show that by comparing it.

Lean thinking applies here. Try removing variability and waste before adding capacity. A change that makes processing times more consistent can beat a new hire, and a model can show you whether it does.

Scenario or study?

Quodsi works out which you're making from the changes you enter, so you don't choose up front.

Build a what-if

There are two ways in:

• From a result. In Flow Analytics, click Study on a row. Quodsi fills in a capacity change for that row's constraint. It uses the range declared on the model if there is one. Otherwise it tries from one above the current value up to double.
• From scratch. On the What-if Runs tab, click New what-if.

The builder has two layouts, which you switch at the top. Sentence reads like a question: pick a step, pick what to change, give the values. Table shows every scenario as a row, against the Baseline, so you can edit them directly.

Add a description of the question this what-if answers. You'll be glad of it three weeks later.

When you change more than one thing, Quodsi asks How should these changes combine?

Levers: changes you make often

A lever is a setting you expect to try again and again, with a range you've agreed, such as "Check-in desks, 1 to 4". Set levers on the Levers tab of an activity, generator, connector or resource, all in the Basic view. The Model editor's own Levers tab needs the Intermediate view. Levers appear as one-click cards in New what-if, and a declared range also fills in the study you start from Flow Analytics. See Levers.

Run it

1. Click Create study or Create scenario. The button shows which one you're making.
2. Click Run study and confirm with Run. Tick Include animation if the run window offers it and you'll want to show this scenario to someone.
3. Wait for each scenario to finish. Its status shows on its row.

Read the comparison

On the Results tab, Overview shows each headline number with its percentage change from the Baseline. On Dashboard, the Comparing bar lets you add scenarios to the comparison.

With two or more finished scenarios in a study, click View tradeoff. Each dot is a scenario. You choose what goes on each axis. A good first view puts your main measure up the side and Scenario: Total cost along the bottom, once you have costs in the model. Look for the bend, the point where spending more buys little. That is often your recommendation.

When you read the numbers:

• Check all your measures. Go back to the ones you chose when you framed the problem. A change that improves one and harms another is a real finding.
• Look at the spread as well as the average. A small gain that disappears inside the normal day-to-day variation isn't a gain you can promise. If two scenarios look close, use more replications before you call a winner.
• Check where the constraint went. Fixing one bottleneck often reveals the next one. Open Flow Analytics on your best scenario.
• Test your shakiest assumption. Re-run your best option with the uncertain input at its pessimistic value. If the answer holds, say so. If it flips, your client needs to know.

Reach a finding

Every comparison should end in one of three outcomes:

Write each one down in a sentence or two, with the scenario names and the numbers. If it leads to a new question, go round again.