# Source Strategy This document describes the public-data source posture for [`urban-observatory`](../README.md)'s v0 work. It is prioritized but open-ended: it names source tiers and reliability considerations, not specific canonical datasets. For the broader method, see [`methodology.md`](methodology.md). For the v0 scope this source strategy supports, see [`v0-scope.md`](v0-scope.md). ## Public-data only The project uses only publicly available data — plans, environmental reviews, agendas, permit records, progress reports, capital plans, funding records, policy documents, and other artifacts accessible to any researcher or member of the public. This is a deliberate architectural choice; see [`methodology.md`](methodology.md). The method's value comes from connecting and interpreting public sources, not from acquiring novel data. ## Provenance discipline Every claim the method produces should be inspectable to its source. Each source contributing to a finding or signal card is recorded with: - the source document or dataset - the retrieval date - the relevant location within the source - the confidence in the interpretation derived from it - any limits or caveats the source carries Outputs without provenance are not the same artifact class as outputs with provenance; see [`methodology.md`](methodology.md). ## Two tier axes — keep them distinct This document's **source ladder** (Tier A / Tier B) is a different axis from the **analytical-depth pass** (Tier 1 / Tier 2) used in v0 scoping (see [`v0-scope.md`](v0-scope.md)). The source ladder describes *what kind of reading a source supports* — signal visibility (Tier A) versus dependency interpretation (Tier B). The depth pass describes *how far v0 interprets* — a shallow citywide baseline (Tier 1) versus deeper interpretation of flagged cases (Tier 2). A Tier-A source can be read at Tier-1 or Tier-2 depth; the two axes are independent. ## Tier A — cross-system signal visibility Tier A is the **signal-visibility** layer: shallow, source-linked public-record signals across the systems that bear on implementation, not housing records alone. Tier A states what the record shows — *signal present / absent / unclear · source found / not found · status changed · dependency flagged · public-record silence · procedural gate visible · public-record wall* — and **implies no causation**. Tier A is a **visibility posture, not a source-collection mandate.** It can surface signals across the bearing systems where those signals are shallowly visible in authorized public sources or already-pinned records: - **Housing / planning** — Housing Element site inventory, APR, development pipeline, planning applications, DBI building permits, completions, zoning and parcel GIS *(the sources pinned and exercised in v0)* - **Infrastructure** — capital plan and infrastructure-investment references - **Transportation** — transportation plans and transit-access layers - **Capital / funding** — affordable-housing and public funding / subsidy signals - **Environmental / resilience** — environmental-constraint and resilience layers - **Public / institutional** — public and institutional records v0 operates on the housing / planning sources already pinned (see [`source-inventory.md`](source-inventory.md)); the non-housing systems are named here as the Tier-A visibility posture — surfaced where public records make them shallowly visible — not as a present collection requirement. Specific datasets within each system are not named here as v0-canonical; dataset selection is part of the prototype's own work. ## Tier B — cross-system dependency interpretation Tier B is the **interpretation** layer, triggered when the project must read *how systems interact* or whether a dependency *materially affects* implementation: explanation, causality, delivery-risk claims, intervention priority, cross-system dependency interpretation, and higher-confidence, expert-review-facing, or public-facing claims. Tier B is **gated** — Tier-A visibility never implies a Tier-B interpretation on its own. Tier B is a *function*, not a separate source list. It interprets deeper reads of the same public records (and, where a specific finding, claim, reviewer need, or demo concept requires it, additional sources such as capital / infrastructure references, transportation plans, funding / subsidy records, environmental-constraint layers, construction-cost / interest-rate / vacancy indicators, or policy / program documents). It is not exercised in v0 by default. ## Tier C — selective supporting source categories Tier C sources are used selectively for Tier-2 deeper interpretation of flagged sites. They are not expected to be exhaustively retrieved in v0's Tier-1 shallow pass. Tier C categories include: - Board of Supervisors and Planning Commission agendas and decision records - CEQA notices and environmental impact reports - public comments and meeting transcripts - code enforcement records - ownership fragmentation analysis - litigation and appeals data - utility-capacity documents - neighborhood-specific and area plans Tier C predates the Tier-A / Tier-B visibility-versus-interpretation reframe; its placement on the revised ladder is reserved for a separate source-taxonomy review and is left unchanged here. ## Posture The source strategy is a posture, not a manifest. New sources may emerge during prototype work, and some named categories may turn out to be less useful than expected. The category-level naming is durable; specific dataset choices remain candidate. Tier A names a visibility posture across systems, not a commitment to collect every system's data: v0 exercises the housing / planning sources, and other systems' signals enter only where public records make them shallowly visible. No data is fetched as part of this document; retrieval is a separate, later phase.