Pilot measurement
How to measure the real time saved by workflow automation
Measure the work people still do, including review and corrections. A faster automated step does not establish that the whole workflow improved.
Set the start and finish before timing
Use the same boundary for the existing process and the supervised process. For example, start when an operator begins reviewing a request and finish when the approved record has been checked against the source. Include every person who touches the work.
Separate hands-on time from elapsed time
Hands-on time includes reading, entering data, approving, correcting, and collecting proof. Elapsed time also includes waiting for an external response or the next working day. Report both, but do not present waiting time removed as labor saved.
Use comparable work
Compare requests of similar complexity, completeness, and system requirements. Record the sample size and how cases were selected. Keep failed runs, duplicates, and exceptions in the record; excluding difficult cases can make a pilot look more effective than it is.
| Field | What to record |
|---|---|
| Case and method | Anonymous case identifier, manual or supervised method, complexity category |
| Hands-on minutes | Preparation, execution, approval, correction, and verification time |
| Elapsed minutes | Total duration, with external waiting identified separately |
| Quality and exceptions | Incorrect fields, rework, failures, and escalation reasons |
| Completion evidence | Whether the final result was verified and accepted |
An illustrative calculation
If a manual case requires 60 hands-on minutes and a comparable supervised case requires 30 minutes including review and corrections, the difference is 30 minutes. The reduction is 30 divided by 60, or 50%. These numbers are an unmeasured example, not a MARPBOUND customer result.
For 100 comparable cases, that example would imply 50 hours of capacity released. It would not automatically mean lower payroll costs or 50 hours of additional billable work. Report setup, training, and ongoing support separately when assessing the business case.
Agree on the acceptance decision
Review time savings alongside correctness, exception handling, and evidence coverage. Report the median and range as well as the average where the sample supports it. If the pilot lacks comparable runs, say that the evidence is incomplete rather than extrapolating a confident savings claim.
See the MARPBOUND pilot measurement framework or return to selecting the first workflow.