Put a dollar value on an untested people decision
Every people decision carries two financial exposures: the downside of a poor outcome and the value a rigid rule leaves behind. Nodes is the People Decision Engine. It starts with one decision, replays it against your own outcomes, and compares the cost of action with the cost of inaction. Finance approves the assumptions. The operating team makes the final call.
Start with one bounded decision
A useful Decision Replay has one owner, one repeated decision, one measurable outcome, and enough historical cases to test. In hiring, that might be who advances from screening and whether that person reaches a defined production milestone. The replay uses past data only. No live decision is touched.
Before the work begins, the customer and Nodes define:
- the decision and population in scope
- the outcome finance accepts as measurable
- the source systems and permitted evidence
- the baseline, measurement window, and success threshold
- the stop conditions and human approval path
The commercial unit is the named workflow from trigger through approved action and evidence. Nodes does not charge per seat, user, token, model call, action, or credit. A quote changes when the workflow is added to or materially broadened.
What one production record shows
The public evidence comes from one Fortune 500 insurance carrier. It demonstrates the method in enterprise talent and cannot forecast another company or decision type.
| Finance question | Reference evidence | Limit |
|---|---|---|
| What did one rigid filter put at risk? | $17.7M in annual production across 2,863 producing agents the filter would have rejected | Retrospective counterfactual from one carrier |
| What changed in the live operating record? | $1.58M in Q1 net savings, CFO-validated | Customer-attributed result that does not assign every operational change to Nodes |
| How much did the hiring loop compress? | 127 days to 38 days, requisition to hire | One deployment, with no promise that another program will match it |
| How much sooner did producing hires reach the first production milestone? | 109 days to 62 days to the first production milestone, a 47-day reduction | Median comparison within the reference evidence |
The Decision Traces paper documents the research method. The case study separates observed results from modeled counterfactuals.
Build the unit economics from your own baseline
A regression on the reference carrier's production records associated each day faster to the first production milestone with $54.35 more in annual production per producing hire. The same evidence set includes an approved 30-day planning reference of $1,357 in projected annual production per producing agent. Both figures are modeled references from one role at one company, not causal estimates or realized savings. A Decision Replay tests your own volume, outcome rate, delay cost, and value per outcome before anyone uses either figure in a budget case.
For a hiring workflow, finance can inspect this equation:
eligible decisions × outcome rate × value per outcome × measured improvement
The model should also show the cost of delay, the downside of a weak outcome, and the value hidden by a rigid rule. Run the hiring ROI calculator to see the method with your own inputs.
Three finance questions ready for historical validation
The production evidence above comes from hiring at one insurance carrier. These adjacent people decisions are candidates for a separate Decision Replay. They are not validated production use cases today.
- Compensation offer analysis: connect offer terms, acceptance, and post-hire production to test which bands produced the strongest return.
- Vacancy cost: compare the finance-approved daily cost of an open role with the historical ramp time for successful hires.
- Contractor outcome review: connect a vendor or contractor engagement to a defined delivery outcome and test whether the contracted result arrived.
Each program needs its own outcome, permitted evidence, baseline, policy boundaries, and named human owner. No hiring result transfers automatically.
Frequently asked questions
How do you separate contribution from coincidence?
Every proposal, human decision, approved action, and later outcome is linked in a signed Decision Trace. Finance sets the baseline and attribution rules before the workflow begins. Where the data cannot support attribution, the business case should say so.
What if the carrier result does not transfer?
It is not assumed to transfer. Every new company and decision program starts with a separate historical replay and a customer-approved threshold. If the evidence does not clear that threshold, the production scope does not proceed.
Who controls the assumptions?
Your finance team does. Nodes shows the evidence, sensitivity, and payback model. The named business owner approves, edits, or declines each proposed action.
Put one people decision to a 72-hour test
Bring one repeated people decision and at least two years of linked decisions and measured outcomes. Before the clock starts, Nodes checks the volume, join keys, outcome coverage, and data quality. Your finance team defines the $1M evidence bar and calculation in writing.
The 72 hours start only after Nodes accepts the agreed deidentified dataset delivered under a mutual NDA. The Replay is read-only. No live decision or production workflow changes. The result is historical evidence, not realized savings. If the historical record falls short of the written bar, there is no production fee and you keep the analysis.