Skip to main content

Geocoding Auditor

Audits a Tableau workbook's geocoded fields for role-assignment mistakes, unmatched-location rates, and ambiguity that silently break maps. Read-only. Every fix is a Desktop UI right-click instruction.

Example: You point it at a workbook with a "City" field that's actually storing zip codes. G1 Role Mismatch fires (Warning). Also flags 22% of that field's values as Unknown (G2 Critical). Recommends role change to Zip Code and a data-quality follow-up.

What it does​

Runs six locked checks (G1-G6) covering role mismatches, unmatched and ambiguous rates, wrong-country resolution, missing hierarchy, and role-vs-datatype conflicts. Novel domain: no other skill in the portfolio covers spatial-role hygiene.

When to use it​

  • "Geocoding audit."
  • "Why won't my cities match?"
  • "Unknown locations across my map."
  • "Check geographic roles."
  • "Why is my map wrong?"

When NOT to use it​

  • Building a map (use worksheet-builder).
  • Map interaction design (use viz-planner).
  • Performance on high-mark maps (use extract-optimizer or publish-readiness).
  • Recommending which cities/countries to include (business decision).

The Six Checks​

  • G1 Role Mismatch: field's current geographic role doesn't match its actual values (city role on zip data, etc.).
  • G2 High Unmatched Rate: >15% Critical, 5-15% Warning, <5% Info.
  • G3 High Ambiguous Rate: shared names resolving to wrong country, same thresholds.
  • G4 Wrong Country Resolution: "Springfield" resolving to the wrong country because no Country context is present.
  • G5 Missing Hierarchy: Country present but no State/City for drill.
  • G6 Role-vs-Datatype Conflict: geographic role assigned to a Number field or vice versa.

What you get back​

  • Severity-ranked findings grouped Critical → Warning → Note → Info.
  • Each finding: check ID, field, current role, recommended role, evidence, Desktop UI fix.
  • Handoff line if the follow-up is map-building or performance tuning.

Limitations​

  • Cannot see rendered map output. Findings inferred from field metadata plus sampling.
  • Cannot fix a role assignment. Right-click Desktop UI only.
  • Cannot detect visual map issues (clipping, layer order, background style).
  • No custom-polygon or spatial-file validation.
  • Sample size limited by get-summary-data default; unmatched/ambiguous rates are estimates on very large tables.