What is ROD?
ROD measures how effectively you use decision models.
ROD, Return on Decisions, looks at how you and your code use models like TypeSafe’s Jev: typed state and questions in, typed decisions with probabilities and confidence out.
Decisions, Not Conversations
A chat model writes an answer for you to read. A decision model makes a decision your code can act on, weighing every allowed answer at once.
Someone reads it, then decides.
- Approve86%
- Escalate11%
- Deny3%
Confidence High
Your code acts on it.
What ROD Looks At
How a decision is built, and what it returns.
State
Question
"Approve the refund?"
- Approve86%
- Escalate11%
- Deny3%
How It’s Built
What It Returns
Confidence, Put to Work
A decision model says how sure it is. ROD looks at whether your code listens.
- Refund #48213ApproveHighActed on
- Refund #48214DenyHighActed on
- Refund #48215ApproveLowEscalated to you
Every decision it accepted held up.
- Refund #48213ApproveHighActed on
- Refund #48214DenyHighActed on
- Refund #48215ApproveLowActed on anyway
One accepted decision didn’t hold up.
Built for Every Decision Model
ROD measures the decision-model category, not one vendor’s model, so results stay comparable as new models arrive.
Jev
TypeSafe AI
Clef
Cloudflare
SubQ Snap
Subquadratic
ROD
One Measure
See how well you use decision models.
Connect Worsley and keep building. Worsley measures how well your decisions pay off and tells you what to change.
- State
- Refund request, order #48213
- Question
- Approve the refund?
- Approve82%
- Escalate13%
- Deny5%