Context Stability
Many-to-Many Relationships and AI: What Can Go Wrong
Many‑to‑many relationships can produce unexpected filter behavior for AI queries.
Rendering diagram...
Dimension → bridge → fact path.
TL;DR
- • Many‑to‑many relationships amplify ambiguity.
- • Bridge tables create safer, deterministic paths.
The problem (layman)
- • Multiple records on each side create unclear filter paths.
- • AI can’t reason about which path is intended.
Why it matters
- • Incorrect context leads to wrong answers.
- • Explanations may cite the wrong segments.
Symptoms
- • Totals vary when grouping by related dimensions.
- • AI answers differ depending on the phrasing.
Root causes
- • Direct many‑to‑many relationships without a bridge.
- • Lack of clarity on the “driving” table.
What good looks like
- • Bridge tables with clear relationship direction.
- • Filters flow predictably from dimensions to facts.
How to fix (steps)
- • Replace direct many‑to‑many relationships with a bridge.
- • Document the intended filter direction.
- • Test with deterministic queries.
Pitfalls
- • Using bidirectional filters to patch issues.
- • Ignoring effect on AI queries.
Checklist
- • Bridge tables for many‑to‑many cases.
- • Documented filter direction.
- • Context tests pass.