“We saved 20 hours a week” is the kind of number that shows up in a proposal and never gets checked. It usually comes from multiplying a guess about time-per-task by a guess about task volume, with no measurement of what the work actually cost before.
Automation can be genuinely worth doing and still fail to justify a claim like that. Here is how to measure it in a way that holds up when someone in finance pushes back.
Measure the baseline first
You cannot report a reduction you never measured. Before changing anything, time a representative sample of the manual process — enough instances to cover a normal and a busy day. Write down who does it, how long it takes, and how often it happens. If you skip this, every later number is an estimate wearing a lab coat.
Count what changed, not what got automated
Automating a step that was never the constraint does not speed anything up. If intake was waiting on eligibility checks and you automated appointment reminders, throughput is unchanged. Tie the measurement to the bottleneck you were actually trying to move.
Net out the work that moved
Automation rarely deletes effort; it relocates it. Time saved on data entry can reappear as time spent handling exceptions, reviewing automated actions, or maintaining the rules. Measure the new work too, and report the net. A change that shifts 15 hours of entry into 4 hours of exception handling is a good result — but the honest number is 11, not 15.
Tie it to something finance already tracks
“Tasks automated” is not an outcome. Translate the effect into a metric that already appears on a report:
- Staff hours redeployed to work that was being deferred
- Visit throughput, or time-to-schedule
- Days in accounts receivable
- Denial rate, or clean-claim rate
If the automation doesn't move one of these, that is useful to know before you scale it.
Give it a fair comparison window
Compare the period after the change against a comparable period before it — same season, same staffing, similar volume — not against the worst week that motivated the project. Seasonality and one-off events will otherwise do the work your automation was supposed to.
The honest number is smaller and more durable
A measured result is usually less impressive than the pitch and far easier to defend a year later. Until you have a real before-and-after, the right thing to put in a plan is a labelled placeholder — [METRIC TO CONFIRM] — and a commitment to measure it, not a number nobody can trace.