sspl-direct-sql-both-publishers

Date: 2026-09-13. Status: superseded.

Decision. Default SQL takes postcode and integer analysis year. Use the exact matched SSPL record's geography, never redirect large users or resolve splits again. Map the year to SIMD using PHS v3.5 Table 4 and return all 14 ingested PHS/Government measures, distinctly labelled, with the selected data-zone vintage and PHS local-band geography codes. Keep every input row and report postcode/year/missing-value statuses independently of address warnings. PO boxes, unlinked large users and deleted records retain source values; residence eligibility and publication choices are downstream. Preserve the earlier SPD linked-small-user examples under docs/sql/history, not as a default fallback.

Reason. The user chose SSPL's published allocation and requested both publishers so later analysis can choose measures. Re-following links overrides that allocation: in SSPL 2026/2 it changed 147 PHS 2020v2 quintiles, including 50 live postcodes. NRS supplies postcode geography; PHS supplies the health-year edition policy and population-weighted bands. This is not a claim of exact PHS postcode-lookup equivalence or a pre-filtered residential cohort.

Evidence. NRS SSPL information note; PHS v3.5 sections 3.1-3.4, Table 4 and Appendix A. tests/test_sql_sspl.py checks every reshaped value and PHS geography across the saved editions, all year boundaries, duplicate/unmatched events, and unchanged large-user own values. AB24 2TY retains quintile 5, rather than linked AB24 2TN's quintile 1. SQL Server runtime verification remains separate.