Loading
One moment.
One moment.
Every public number on this site traces to one of four evidence layers. This page maps each statistic to its source, population, period, outcome metric, validation type, and status.
Canonical source for every NODES statistic. If a number anywhere conflicts with this table, this table wins.
Every statistic on this site comes from exactly one of these layers. When two numbers seem to disagree, they usually come from different layers, different populations, or different periods.
The arXiv paper, arXiv:2604.19819. Population: 10,765 agents hired at one Fortune 500 carrier, 10,362 evaluable after tenure censoring. Period: 2022 to 2025. Outcome metric: SNA (first production milestone). Validation type: retrospective observational study on customer records.
Population: 747 NODES-scored agents contracted January to March 2025, each with 9+ months of tenure. Period: contracted January to March 2025, outcomes measured after 9+ months of tenure. Outcome metrics: RPA (sustained production award) and SNA milestones. Validation type: out-of-sample validation against realized production milestones.
Population: the live deployment at the carrier, 900,000+ candidates scored since January 2025 across 215+ locations. Period: January 2025 to present, with the year-one results table covering 2025. Outcome metrics: candidates scored, time-to-hire, screening and interview volume, documented savings. Validation type: observed customer production records.
Population: 1,690 producing agents analyzed, 2022 to 2025, 679 with scoring and 1,011 pre-deployment. Period: 2022 to 2025. Outcome metric: annual production per producing hire. Validation type: regression analysis on production records. This layer yields the $54.35/day constant, with $34.95/day controlling for source channel and tenure.
One row per public statistic. Status reads as follows. Validated means tested against realized outcomes. Observed means measured directly from production records. Pilot observation means from a bounded pilot. Modeled means estimated by a fitted model or a counterfactual.
Each row carries an anchor, /evidence#c01, /evidence#c20, and so on, so other pages can deep-link to the exact claim.
| Claim | Value | Layer | Population | Period | Outcome metric | Validation type | Status |
|---|---|---|---|---|---|---|---|
| C01Candidates scored since January 2025 (cumulative, live deployment) | 900,000+ | L3 | Live deployment, one Fortune 500 carrier | Jan 2025 to present | Candidates scored | Customer production record | Observed |
| C02Candidates scored in year one (2025) | 730,000+ | L3 | Same deployment | 2025 | Candidates scored | Case-study year-one results table | Observed |
| C03Hires via NODES since January 2025 | 7,000+ | L3 | Live deployment, one Fortune 500 carrier | Jan 2025 to present | Hires | Customer production record | Observed |
| C04Contracted hires (2025) | 2,335 | L3 | Same deployment | 2025 | Contracted hires | Case-study year-one results table | Observed |
| C05Study population, agents hired 2022 to 2025 at one Fortune 500 carrier / evaluable after tenure censoring | 10,765 hired / 10,362 evaluable | L1 | Hired agents at one Fortune 500 carrier | 2022 to 2025 | SNA (first production milestone) | Retrospective observational study, arXiv:2604.19819 | Observed |
| C06Agents hired in 2025 in the published study (research contexts) | 3,623 | L1 | 2025 study cohort | 2025 | SNA (first production milestone) | Retrospective observational study, arXiv:2604.19819 | Observed |
| C07Producing agents analyzed, 2022 to 2025 (679 with scoring, 1,011 pre-deployment) | 1,690 | L4 | Producing agents at the carrier | 2022 to 2025 | Annual production per producing hire | Speed-to-production analysis | Observed |
| C08NODES-scored agents, contracted Jan to Mar 2025, 9+ months tenure | 747 | L2 | Scored agents at the carrier | Contracted Jan to Mar 2025 | SNA and RPA milestones | Out-of-sample validation cohort | Observed |
| C09Of the 747, agents with parseable ATS skills text | 535 | L2 | Subset of the 747 | Contracted Jan to Mar 2025 | RPA milestone | Keyword-ablation subset | Observed |
| C10RPA achievers with ATS skills text, the base for the 80% and 98% filter-elimination claims | n=50 | L2 | Subset of the 535 | Contracted Jan to Mar 2025 | RPA milestone | Funnel base within the out-of-sample cohort | Observed |
| C11Locations | 215+ | L3 | Live deployment footprint | Jan 2025 to present | Operational footprint | Customer production record | Observed |
| C12Applications per year at the carrier | 1.5M | L3 | Carrier hiring funnel | Annual | Application volume | Customer record | Observed |
| C20Documented year-one savings at the carrier | $1.58M | L3 | Live deployment | 2025 | Savings | Customer production record | Observed |
| C21Annual premium credit from 2,863 producing hires the insurance-experience filter would have rejected | $17.7M | L1 | 2,863 producing hires | 2022 to 2025 | Annual premium credit | Retrospective counterfactual on observed production, arXiv:2604.19819 | Observed |
| C22Producing hires the insurance-experience filter would have rejected (pairs with C21) | 2,863 | L1 | Study population | 2022 to 2025 | SNA production | Retrospective counterfactual on observed production, arXiv:2604.19819 | Observed |
| C23Each day faster from contract to first production milestone equals $54.35 more annual production per producing hire | $54.35/day | L4 | 679 producing agents with scoring | 2022 to 2025 | Annual production per producing hire | Linear regression on production records | Modeled |
| C24Speed constant, conservative: controlling for source channel and tenure | $34.95/day | L4 | 679 producing agents with scoring | 2022 to 2025 | Annual production per producing hire | Regression with controls for source channel and tenure | Modeled |
| C25Ramp acceleration, median days from contract to first production milestone, 2025 scored cohort vs pre-deployment | 62 vs 109 days, 47 days faster | L4 | 2025 scored cohort vs pre-deployment baseline | 2022 to 2025 | Days to first production milestone | Median comparison of production records | Observed |
| C26Time-to-hire, requisition to hire. A different metric from C25, never blended with it. | 127 to 38 days (70% reduction) | L3 | Live deployment | 2025 | Days from requisition to hire | Customer production record, year-one results table | Observed |
| C28Top-performer multiplier. Hires scoring 72 or above achieve top-performer milestones at 2.47x the rate of those below 72. Never "written premium, scored vs unscored." | 2.47x | L2 | 747 NODES-scored agents | Contracted Jan to Mar 2025 | Top-performer milestone rate | Out-of-sample, n=747, p=0.006 | Validated |
| C30Combined-signal gap: RPA rate 15.3% vs 2.9%, sales keyword plus NODES score 72 or above vs no sales keyword plus score below 72 | 5.37x | L2 | Segment within n=747 | Contracted Jan to Mar 2025 | RPA (sustained production award) | Segment analysis within n=747, p=0.0007 | Validated |
| C31Quintile calibration: SNA 15.9% to 22.5%, RPA 5.1% to 13.0% | 2.55x Q1 to Q5 | L2 | n=747 quintiles | Contracted Jan to Mar 2025 | SNA and RPA rates by score quintile | Quintile calibration on the out-of-sample cohort | Validated |
| C32Legal approval after six vendors were rejected over 18 months at the same carrier | 17 days | L3 | The carrier's vendor review | At deployment, live since Jan 2025 | Approval time | Customer record | Observed |
| C33Deployment, contract to production | 34 days | L3 | The carrier's deployment | At deployment, live since Jan 2025 | Days from contract to production | Customer record | Observed |
| C34Screening reduction | 40% fewer manual screens | L3 | Live deployment | 2025 | Manual screens | Customer production record, year-one results table | Observed |
| C35Interview reduction | 2 fewer interviews per hire | L3 | Live deployment | 2025 | Interviews per hire | Customer production record, year-one results table | Observed |
| C40ATS keywords parsed / testable | 8,181 / 3,597 | L1 | Applicant pool at the carrier | 2022 to 2025 | Keyword testability | Keyword census, arXiv:2604.19819 | Observed |
| C41Predictive keywords after Bonferroni correction | 0 | L1 | 3,597 testable keywords | 2022 to 2025 | SNA production | Hypothesis testing with Bonferroni correction | Validated |
| C42Anti-predictive keywords (median OR 0.749, 70.2% directionally negative) | 30 | L1 | 3,597 testable keywords | 2022 to 2025 | SNA production | Hypothesis testing with Bonferroni correction | Validated |
| C43License finding, SNA production rate with vs without a license | 24.9% with vs 33.1% without | L1 | Study population | 2022 to 2025 | SNA production rate | Odds-ratio analysis, arXiv:2604.19819 | Validated |
| C44Insurance-experience finding: anti-predictive on SNA. On RPA (n=535) the same flag is null (p=0.56). The two outcomes never mix in one claim. | 28.0% with vs 33.7% without, OR 0.763, p<0.001 | L1 | Study population, RPA check on n=535 | 2022 to 2025 | SNA production rate, RPA | Odds-ratio analysis, p<0.001 on SNA | Validated |
| C45Zero-keyword high scorers: agents with no traditional keywords produced at 33.7% vs 26.3% for all-keyword agents | 677 agents | L1 | Study population | 2022 to 2025 | SNA production rate | Observed production-rate comparison, arXiv:2604.19819 | Observed |
| C46AUC, ATS keywords (research cohort) | 0.558 | L1 | n=7,055 | 2022 to 2025 | SNA production | AUC on the research cohort | Validated |
| C47AUC, PI standalone | 0.647 | L1 | n=229, small research sample | 2022 to 2025 | SNA production | Cross-validated AUC | Validated |
| C48AUC, full fusion. Carries the paper's small-sample caveat: measured on the 229 agents with personality data, a research-sample result. | 0.735 | L1 | n=229, small research sample | 2022 to 2025 | SNA production | Cross-validated AUC on n=229 | Validated |
| C49AUC, deployed score (initial_fit_score) | 0.57 | L1 | n=714 | 2022 to 2025 | SNA production | AUC on the deployed score | Validated |
| C50AUC out-of-sample, NODES score vs six-keyword composite (p=0.006). The preferred pair for marketing surfaces. | 0.618 RPA / 0.567 SNA vs 0.548 RPA / 0.512 SNA, combined 0.644 | L2 | n=747 | Contracted Jan to Mar 2025, 9+ months tenure | SNA and RPA | Out-of-sample AUC, p=0.006 | Validated |
| C51Behavioral-score moderation, high vs low score bands. Suggestive pending larger samples. | $114/day vs $41/day (2.8x) | L1 | Score bands within the study population | 2022 to 2025 | Production per day of ramp acceleration | Moderation estimate, suggestive pending larger samples | Modeled |
| C52Skills-count finding: candidates with 0 to 10 parsed skills produced at 36.1% vs 22.7% for 51+ skills | 0 to 10 skills: 36.1% vs 51+: 22.7% (r = -0.090) | L1 | Study population | 2022 to 2025 | SNA production rate | Correlation, r = -0.090 | Validated |
Rows C27 and C29 are reserved. Those figures do not currently appear on the public site. The study behind L1 and L2 is on the case study, and field findings from the paper are on research.