Calculator field guide / 01
Check the model before you hand over the tool.
A useful calculator must do two things: implement its formula correctly and represent the task you actually need to plan. Example checks help with the first. Clear assumptions and independent review help with the second.
Start with an independently known answer
Choose a small case you can work out without the tool. Write the expected answer first, then run the calculator. Copying its result into the expected-answer field only confirms that the same calculation can repeat itself.
For a serial process with 30 minutes of active work and 90 minutes of waiting, total lead time is 120 minutes. Active work occupies 30 / 120 = 25% of that time.
This example assumes the steps do not overlap and all values refer to one item. A batch process with parallel work needs a different model.
Choose rounding for the decision
Rounding changes what an operational answer means. If 51 learners need seats in cohorts of 20, 51 / 20 is 2.55. You need 3 cohorts, so use ceil(learners / seats). Rounding down to 2 leaves 11 people without seats.
Capacity answers often need the opposite treatment. A team with time for 12.8 whole tasks can finish 12 complete tasks under that model, so use floor(...). Do not apply the same rounding rule to every output.
Test the edges of the allowed range
- Try a zero where zero is allowed. Does the answer still make sense?
- Try the minimum and maximum permitted values.
- Try a value just above a rounding boundary, such as 21 learners in cohorts of 20.
- Check any denominator. If it can become zero, narrow the input range or change the model.
- Include a case that would expose a unit mistake: hours versus minutes, or 75 versus 0.75 for a percentage.
HTML Workflows blocks inputs outside their declared bounds and reports division by zero. Those checks cannot choose the correct limits for your business process.
For a complete numerical example, use the team capacity formula and worked scenarios.
Write down what is excluded
A team-capacity model may assume interchangeable staff and uniform task duration. A training model may exclude preparation and travel. State those exclusions beside the result so a client does not mistake a limited estimate for a complete schedule.
Save the version you actually checked
In the studio, save after the expected-result checks pass. Export requires a saved snapshot matching the current tool and brand. Changing a label, formula, assumption or test case requires another saved version before export.
The saved history is local to that browser. Keep the exported file with your client handoff notes; clearing site data removes the authoring history. The export includes its formula version and example checks, but it is not independent certification of the model.