Trust

Can this recommendation be used?

This is the proof page: data freshness, caveats, claim evidence, model logic, and generated-file audit.

Audit

pass

Start / Trust

National data trust

3 source caveats active. Open Trust before using this in a decision meeting.

NOAA NCDC / NCEI Climate Data Online: cached-validFRED: missingEIA: missing

Source trust

Public data freshness and usability

`cached-valid` means the latest fetch failed but a prior public payload is still used and disclosed. Missing sources are not allowed to create unsupported claims.

Live 4 Cached-valid 1 Stale 0 Missing 2

NOAA / National Weather Service

nws

live

Used: 2026-10-29T13:49:46.022Z

public/data/live/nws.json

NOAA NCDC / NCEI Climate Data Online

ncdc

cached-valid

Used: 2026-10-29T13:49:30.339Z

public/data/live/ncdc.json

U.S. Census Bureau

census

live

Used: 2026-10-29T13:49:46.230Z

public/data/live/census.json

FRED

fred

missing

Used: 2026-10-29T13:49:46.308Z

BLS

bls

live

Used: 2026-10-29T13:49:46.308Z

public/data/live/bls.json

EIA

eia

missing

Used: 2026-10-29T13:49:46.764Z

EPA ECHO

epa-echo

live

Used: 2026-10-29T13:50:00.249Z

public/data/live/epa-echo.json

Can I use this?

Use with caveats

3 source caveats active. Open Trust before using this in a decision meeting.

live

4

cached valid

1

missing

2

Model audit

statistical-risk-v2

risk = 0.30 weather + 0.22 demand + 0.15 facility + 0.12 cost + 0.12 momentum + 0.09 regional concentration - confidence penalty

weather Pressure

30%

construction Demand

22%

facility Context

15%

cost Pressure

12%

score Momentum

12%

regional Concentration

9%

V2 scoring proof

How the model keeps outliers from dominating

Normalization

Raw public signals are converted into comparable 0-100 component scores.

Winsorization

Values are capped at the 5th and 95th percentiles before scoring so one extreme public datapoint does not control the whole recommendation.

Confidence penalty

Source freshness is separate from the raw score. Cached, stale, or missing sources reduce confidence only when that source contributes to the state score.

Action-band thresholds

Why a state is labeled Act, Investigate, Watch, or Monitor

Act

Score 75+, top action band

Investigate

Score 58+, next band: Act

Watch

Score 42+, next band: Investigate

Monitor

Score 0+, next band: Watch

Near-threshold warnings appear when a state is close enough to an action-band boundary that small public-data changes could alter the label.

Component distributions

What range each public signal used in this build

These ranges come from the generated model output and make the scoring math auditable without reading the scripts.

Weather

30%

Median

0

P10

0

P90

0

Construction demand

22%

Median

1487

P10

0

P90

0

Facility/compliance

15%

Median

0

P10

0

P90

0

Cost pressure

12%

Median

1

P10

0

P90

0

Momentum

12%

Median

24

P10

0

P90

0

Regional concentration

9%

Median

84

P10

0

P90

0

Data caveats

3 source caveats active. Open Trust before using this in a decision meeting.

NOAA NCDC / NCEI Climate Data Online

NOAA NCDC / NCEI Climate Data Online is cached-valid: the latest fetch failed, but a prior public payload is disclosed and still used.

FRED

FRED is missing: it should not support unsupported live-data claims.

EIA

EIA is missing: it should not support unsupported live-data claims.

Claim evidence

Narrative claims are tied to generated public data

Current posture

decision-summarystate-signals

Demand context

censusanalytics

Cost movement

blsoperating-model

Public facility/compliance context

epa-echooperating-modelanalytics

Source trust

live-bundleaudit-manifest

Model cards

How the analytics should be trusted

State action score

Rank states for operating review using public signals.

Formula

risk = 0.30 weather + 0.22 demand + 0.15 facility + 0.12 cost + 0.12 momentum + 0.09 regional concentration - confidence penalty

Inputs

  • NOAA/NWS active alerts
  • BLS cost movement
  • public facility/compliance context
  • regional modifier

Limitations

  • State-level triage only
  • Not a facility-level forecast
  • Public-data latency varies by source

Source health

Show whether the current build is relying on live or cached public data.

Formula

live provider count, cached provider count, and audit manifest hashes

Inputs

  • live-bundle.json
  • audit-manifest.json

Limitations

  • A live connection does not guarantee the source has new observations
  • cached-valid means the latest fetch failed but a prior public payload is still being used transparently

BLS cost pressure

Show latest public producer-price movement for tracked cost series.

Formula

latest BLS point versus prior point and trailing series direction

Inputs

  • BLS public API time-series data

Limitations

  • BLS series are proxies and may not map one-to-one to resin or wood-panel costs

NCDC climate probe

Show recent public station observations for weather context.

Formula

recent daily TMAX/TMIN/PRCP values from a documented station

Inputs

  • NOAA NCDC / NCEI GHCND station observations

Limitations

  • Single-station probe until regional station coverage is expanded

Public facility/compliance context

Add public industrial footprint context for priority states.

Formula

public chemical-sector facility rows and map clusters by state, capped in the state score

Inputs

  • EPA ECHO public get_facilities
  • EPA ECHO public get_map

Limitations

  • Clusters are public compliance-context signals, not production capacity or company exposure

Census construction-demand context

Show where public building-permit units may indicate downstream construction activity relevant to wood panels and adhesives.

Formula

state permit units normalized against the current public BPS state file, capped as a small score component

Inputs

  • U.S. Census Building Permits Survey state monthly file

Limitations

  • Permits are demand context, not guaranteed starts, completions, resin demand, or company sales

Generated files

Audit manifest

Open raw generated-file manifest

registry

7

fixtures/source-registry.json

sources

6

fixtures/sources.json

marketBrief

fixtures/marketBrief.json

signals

fixtures/signals.json

stateMap

51

fixtures/us-states-map.json

stateSignals

51

public/data/state-signals.json

decisionSummary

public/data/decision-summary.json

stateGeometry

51

public/data/state-geometry.geojson

operatingModel

51

public/data/operating-model.json

historyIndex

22

public/data/history/index.json

latestSnapshot

51

public/data/history/latest.json

changeBrief

10

public/data/change-brief.json

aiInsights

4

public/data/ai-insights.json

liveBundle

7

public/data/live-bundle.json

valueBrief

5

public/data/value-brief.json

decisionAnalytics

51

public/data/decision-analytics.json

reportView

51

public/data/report-view.json

decisionTree

261

public/data/decision-tree.json

statisticalRiskModel

51

public/data/statistical-risk-model.json