Data Smells Summary

Data Smells Summary

Convert raw diagnostics into operator judgment.

Inputs

• Output from one or more of:
  • schema-smells
  • project-audit
  • spatial-index
• Optional: intended downstream work (publish? overlay? joins? routing?)

Output (format)

1) Plain-language risk summary

• 5 to 10 sentences max.
• Emphasize what breaks and what slows down.

2) Ranked “most likely to bite you later” list

For each item:

• Why it matters (impact)
• When it shows up (trigger: publishing, joins, long runs, editing)
• How to confirm (quick check)

3) Suggested remediation order

• Group by dependency:
  1. correctness blockers (projection, geometry validity, IDs)
  2. reliability blockers (null spikes, inconsistent types, domains)
  3. performance blockers (indexes, geometry complexity)
• Keep the plan short and actionable.

Heuristics

• Prefer correctness over performance.
• Prefer fixes that reduce rework (projection, ID hygiene, geometry validity).
• Call out “debt interest”: issues that compound (domainless coded values, mixed types, unindexed joins).

Example

Input: schema-smells finds nullable IDs + mixed field types; spatial-index flags missing index on join key.

Output: A narrative summary + top 5 risks + a 1-2-3 remediation order.