How we measure results
Any effect figure is the difference between what happened and what would have happened otherwise. Nobody observes the second, so what matters is not the number but the way it was obtained.
How we measure
Locations where the change was not switched on serve as a control. We take the difference in differences: how far the treated group moved beyond how far the control moved.
If the control group is too small, the period too short, or the data had an outage, the system shows “not enough data” rather than a pretty number.
Share of orders a supplier answered; days of coverage; money in dead stock; the number of items that dropped out of a calculation. Quantities from logs, not estimates.
What we do not publish
- We do not publish “85–92% forecast accuracy”: without a stated baseline it means nothing.
- We do not quote dollar savings: clients use different currencies and different ways of computing cost.
- We never present a model calculation as a client result. If it is an example, it says so.
Show us one critical process. We will show how it runs here.
We look at your cycle: how an order is assembled today, who decides, where time leaks and what the system takes over.