Demand taxonomy

Evidence class has three values:

• modeled_gap for curated case needs
• observed_gap for real analyses without a clean market fit
• explicit_request for direct requests for a missing market

The taxonomy response also states the minimum partner group size. Customer counts smaller than that group are omitted from partner output.

Risk types

The current v1 risk vocabulary covers energy, food and agriculture, metals and materials, technology inputs, weather disruption, macro and local demand, financing rates, foreign exchange, promotions, supplier concentration, policy costs, equity markets, crypto prices, elections and an unmapped class. These bounded values let partners compare records without receiving raw owner language.

unmapped is a valid honest result. It preserves a clear need that does not fit the current bounded categories without forcing it into a generic macro type.

Harm direction

Harm direction describes the owner's adverse world, not the contract side. Values cover cost increases, revenue decreases, availability decreases, rate increases, price moves, event occurrence, event non-occurrence, two-sided exposure and unspecified direction. The analyst still reasons from that direction to the exact contract's YES or NO side.

Time horizon

Time horizons use broad buckets: 0 to 30 days, 31 to 90 days, 4 to 12 months, over 12 months and unspecified. These dimensions support aggregation. The analyst uses exact contract dates when judging timing.

Coarse context

Business segment and geography are deliberately broad. Geography is national or regional rather than a street address. Amounts use bands rather than customer monetary values. These fields help market design while reducing the chance that a record points back to one owner.

Runtime source of truth

The authenticated taxonomy resource is the runtime source of truth for exact values and labels across risk type, evidence class, harm direction, time horizon, business segment, coarse geography, at-risk amount band and lifecycle status. Clients should tolerate additive options and should not infer new values from display copy.