Test Incentive Formulas
Testing a formula takes a minute and prevents the worst conversation in a workshop: telling somebody the number on their payslip was a mistake.
Who can do this: Account Owners and Org Admins.
Testing
Open the scheme you are building or editing
Choose Test with Sample Data
Enter figures
Use a real month you remember — a typical technician’s billable hours, revenue, efficiency.
Read the computed incentive
Change the figures and read it again
What to test beyond the average
The average month is the case least likely to go wrong. Test the edges:
| Case | What you are checking |
|---|---|
| An excellent month | That the payout is affordable. Formulas with a percentage and a bonus can compound faster than expected |
| A poor month | That it does not go negative or produce something absurd |
| Just under a threshold | The technician who misses an 80 per cent target by one point |
| Just over a threshold | That crossing it is a sensible step, not a cliff |
| Zero activity | Somebody off sick for the period should produce zero, not an error |
| A very high single figure | One enormous job should not pay a year’s bonus |
Threshold cliffs are the commonest design fault. A bonus that pays nothing at 79 per cent efficiency and a large amount at 80 gives a technician a powerful reason to influence the number rather than the work. Test both sides of every threshold.
Reading the preview
The formula preview shows the formula and the variables it uses. Check that list before you check the number: a formula that does not reference the variable you care about is not measuring what you think, however plausible its output looks.
After it goes live
Compare the first period’s computed figures against what you expected before anybody is paid on them. A scheme that behaves differently on real data than on your test figures usually means the test figures were more optimistic than the shop.
Common questions
The number looks wrong. Check the variables in the preview first, then the thresholds. Most surprises are a threshold applying where you did not intend.
Can I test against real data instead of sample figures? Testing is against figures you enter. Use a real period’s numbers from analytics.
The AI-generated formula tests fine but I cannot follow it. Do not ship it. A formula nobody in the business can explain is a liability at the first dispute, whatever the test says.
Can I save a scheme without testing it? Yes. It is not enforced. It is still worth the minute.