United States — consumer credit
Top drivers
⌁ mcp.call("adw-015") vADW-015-live-1.0 How stressed is US consumer credit right now — rising delinquencies or credit binging?
United States — consumer credit
Top drivers
⌁ mcp.call("adw-015") vADW-015-live-1.0 A credit-monitoring agent polls ADW-015 each month and fires when the score crosses 58 (above its historical mean of 48.8) on a rising trend with confidence > 0.8. The IOM's top_drivers field tells the agent which component is leading — delinquency rate (DRCCLACBS) or revolving credit growth (REVOLSL) — so it can route the alert to the correct downstream workflow: a delinquency spike triggers tightened underwriting rules, while a revolving-credit surge triggers a 'credit binging' flag for collections pre-staging. The source_lineage pointing to FRED and a versioned methodology (z_delinq 0.6 + z_revolving_yoy 0.4, 36-month window) means the action is fully auditable against regulatory exam questions. At the current reading of 54.8 — the 79th percentile in 10 years of history — the agent would already be in an elevated-watch state despite the falling trend, warranting continued surveillance rather than full alert dismissal.
A consumer-bank Chief Risk Officer uses ADW-015 as a monthly forward-indicator layer on top of internal delinquency reports, which lag by 30-60 days due to reporting cycles. When the composite z-score climbs into the upper quartile (as it sits now at the 79th percentile), the CRO convenes the credit committee to review line-of-credit exposure and tighten approval cutoffs for new revolving accounts — a decision that previously required assembling delinquency, charge-off, and revolving-balance data from three separate systems manually. The packaged IOM replaces two hours of monthly data assembly with a single API call whose top_drivers field directs attention to whichever signal is driving stress.
z_delinq (0.6) + z_revolving_yoy (0.4); both z-scored vs trailing 36-period window → composite z → 0-100 (50=neutral, >50=stress rising)
Version ADW-015-live-1.0 · validated to beat a naive baseline · benchmark: none packaged