concept · concepts/System of Record.md

System of Record

status: active · created 2026-06-26 · 1 source · #data-engineering #identity #governance #authority #craft

The authoritative source for a given fact — the one a conflict defers to. The governing caution: a convenient source isn't automatically a system of record. Verify authority before you promote a source to source-of-truth (SOT).

Explanation

A system of record (SOR / SOT) is whichever source is authoritative for an entity or fact — when two stores disagree, this is the one that wins. The source (Working Heuristics — Data, Matching, Product, Org, Tooling) warns against the common slip: treating whatever is easiest to query as if it were authoritative. Convenience (a handy export, a denormalized view, a cache everyone already connects to) is not authority. Promoting a convenient source to SOT without checking who actually owns and governs the fact bakes a quiet error into everything downstream.

Verify authority first: who produces the fact, who is accountable for its correctness, where it's edited. That's the SOR — which may well be a less convenient place than the one people reach for.

Why it matters / where it shows up

  • Authority errors are silent and compounding. If a non-authoritative source is trusted as SOT, every reconciliation and match downstream inherits its drift, and nobody sees the moment it went wrong.
  • It anchors identity resolution. After matching links records, the SOR decides which attributes are canonical.
  • It rhymes with reading lineage honestly. "Verify authority before promoting to SOT" is the identity-layer version of [[Pipeline Legibility|"trust lineage you can read; don't assert what you can't see"]] — and the [[Lands vs Lives|table data lands in]] is often the real SOR, not the convenient view it lives in.

Open questions

  • When no single source is fully authoritative, how do you compose a SOT from several partial authorities without re-introducing the "convenient source" trap?

Connections

  • Shrink the Candidate Space — matching generates candidate links; the SOR adjudicates which record is canonical.
  • Pipeline Legibility — both insist on verifying what's actually authoritative vs. asserting convenience.
  • Lands vs Lives — the physical table data lands in is frequently the true SOR.

Sources

The wiki is the source. This door reads it at build time and never writes to it.