Urd Atlas methodology
Limitations & Boundaries
This page separates disclosure boundaries from analytical interpretation boundaries, so customers can see both what the public trust layer does not reveal and what the product should not be used to infer.
What the public methodology discloses
- Field meaning and artifact ownership
- Interpretation of confidence, scorecard, regime, and freshness
- Publicly relevant thresholds and gates
- Worked examples for deterministic and gated outputs
What the public methodology does not disclose
- Exact upstream AWS schemas or join logic
- Intermediate feature tables and ingestion repair rules
- Enough implementation detail to reconstruct raw source rows
- Enough implementation detail to clone the full private pipeline
Analytical interpretation boundaries
- Labels and scores are chain-relative. A
CONGESTEDlabel on Arbitrum says nothing by itself about Ethereum. - Labels can change day to day.
CONGESTEDandCHEAPcan be triggered by single-day threshold crossings, whileHEATINGdepends in part on a trend condition. - Customers who require multi-day regime stability must apply their own duration filter or smoothing rule downstream.
regime.drivers[].z_robustand scorecard dimension scores are not expected to be numerically identical, because they use different input series, windows, and purposes.- BTC capacity is an instability proxy around the recent block-time norm. It does not distinguish directional “fast” versus “slow” block-time states in the way a raw block time field would.
- Friction can be elevated because fee burden relative to transferred value is elevated, not only because native-denominated fee size is elevated.
What this means for downstream analytics
Urd Atlas is designed to publish descriptive, chain-relative analytical states. It is not designed to eliminate all label volatility, provide cross-chain absolute rankings, or serve as a complete substitute for user-defined downstream smoothing and segmentation rules.