How to measure what automation actually changed
Three numbers that tell you, one quarter in, whether the project paid off.
Also inRU
"We saved time" is not a measurement
A client asks to automate a process and expects things to be better afterwards. Nobody measures "better": before the rollout no one counted the hours, and after it there is nothing left to compare against.
We record three numbers before the work starts. Not ten — a person will agree to collect three.
What we record
Cycle time. How long it takes from the triggering event to a closed task: order to shipment, shift to a signed timesheet. Measured by hand, across the last ten cases, before any work begins.
Share of manual entry. How many fields a person copies from one system into another. This is the one number that moves sharply and predictably once an integration lands.
Cost of an error. What a single typo in a delivery note costs: a return, a recount, a fine. It usually turns out to exceed the cost of the operation itself.
Reading the numbers a quarter later
The same three measurements, on the same ten cases. If cycle time held steady while manual entry dropped, the automation moved the work rather than removing it. That is still a result — but it cannot be sold as savings.