Every AI vendor on the market will hand you a number. It'll be large. It'll be flattering. And it won't survive a single pass of due diligence from a competent finance team. We know because we've watched it happen, dozens of times.
The core problem is that most ROI numbers are built from a single survey of self-reported productivity lift, projected across an organization, and multiplied by a loaded fully-burdened cost figure. That's a persuasive narrative. It isn't a measurement.
Nymbral's 0–150 ROI Score starts somewhere else. It starts with the contract: what you actually pay, seat-by-seat. Then it layers adoption telemetry (who logged in, who used the feature, who abandoned it), task-level automation baselines from published research, and a conservative impact envelope that accounts for uncertainty.
The output is a number you can defend, not a number that sells. That distinction, between a marketing metric and a measurable one, is what separates organizations that are operationally AI-native from organizations that are paying for the appearance of it.
If your AI portfolio can't answer the simple question 'what did we get?' in terms that trace back to data your auditor could verify, you don't have an AI strategy. You have a line item.