1. Start with the user's decision
A feature name does not make the problem clear. Ask what the user needs to decide, under what circumstances, and where they face difficulty today.
Whose decisions and actions need to improve, and how?
“The dispatcher reviews and confirms a route plan that incorporates new orders” reveals the work more clearly than “display a map.”
Calculating the plan is only part of the task. Consider the full workflow: check the inputs, compare results, make changes when needed, and confirm the final plan.
2. Separate objectives from mandatory constraints
An objective describes what you want to improve. A constraint describes a condition you must satisfy. Reducing travel time and keeping each vehicle within its capacity play different roles.
| Question | Fictional example |
|---|---|
| What needs to be decided? | The vehicle assigned to each order and the sequence of visits |
| What should improve? | Reduce total travel time while meeting the agreed conditions |
| What must be satisfied? | Vehicle capacity and visit time windows |
| What data is required? | Order locations, quantities, and time windows; vehicle information; travel times |
Do not assume every requirement can be supported at once. Check whether the computational service supports the chosen objectives and constraints, and whether the required data is available.
3. Test a key assumption with a small experiment
Once the problem is clear, test an assumption that could prevent the solution from working before building the full system. For example, use fictional orders to ask: “Can the available data produce a plan that a dispatcher can meaningfully review?”
Compare the existing and proposed approaches under the same conditions, and record what still needs to be checked. A small experiment does not establish success across a broader scope.
Try it yourself
Choose a familiar task and complete these four sentences. Replace sensitive business information with fictional values.
- The user is ______ and needs to decide ______.
- The result that should improve is ______.
- The mandatory constraints and required data are ______.
- The first assumption to test, and how to test it, are ______.
Review your answers
Check that your description shows the user's work rather than just a feature name, and separates improvement objectives from mandatory conditions. You are ready for the next step when you can explain which result would lead you to proceed or revise each assumption.