Digital Twin Series · Level 4 · Optimisation
What-If Analysis and Scenario Simulation
What happens to output, WIP and cost if we change speed, staffing, product mix, machines or buffers, and which change is worth making?
A what-if analysis runs the question on the twin instead of on the shop floor. You change one or more inputs, run the simulation, and compare the results with the baseline: throughput, lead time, work in progress, bottleneck, utilisation and cost. This page has a working five-station line simulator. Build your own scenarios, run eight example scenarios, see where the bottleneck moves, find which lever matters, and see how certain the answer is.
| Parameter | Typical what-if options | In the simulator below |
|---|
| Scenario | Throughput (parts/h) | Daily output | Lead time | WIP | Bottleneck | Cost per part | Demand |
|---|
The line has five stations: Load, Assembly, Weld, Test and Pack. Assembly and Test are manual, the others are machines. Each result is the average of six simulation runs with different random seeds, over a 3 hour run after a 10 minute warm-up. Move the sliders, then save up to three scenarios and compare them with the baseline. The relationships between levers and cycle times are assumptions of this teaching model (an extra operator cuts a manual cycle time by 70% of the first, speed above 100% adds variation, parallel machines share the work). Replace them with your own data.
| Scenario | Description | Throughput (parts/h) | Lead time (min) | WIP (parts) | Bottleneck | Line utilisation | Remark |
|---|
Utilisation is the share of time a station is working. The bottleneck is the busiest station, the one that limits the whole line. Orange bars are manual (operator) stations and blue bars are machines. A balanced line has stations at similar, healthy utilisation. Speeding up any station other than the bottleneck changes little.
Each bar is one change made alone to the baseline line, with the others unchanged. It shows how a what-if run finds the changes that pay off and the ones that do not.
Each run uses different random numbers for cycle time variation and breakdowns, so each gives a slightly different answer. A single run can mislead: report the average and the spread, and the chance of meeting the target.
Figures are indicative and the simulator is a simplified teaching model. Build and validate the model of your own line before you decide.