BreezyCorp Handbook

Appearance

Suggestion confidence is all LOW ​

The Mappings tab shows every line item suggestion at LOW confidence; the alternatives[] list on each row is filled with implausible candidates.

What this means ​

The suggestion engine ran (exact-match against mapping_library.json → fuzzy-match against taxonomy element labels) but produced nothing close. Almost always an input problem.

Causes ​

  1. Wrong entry point on the filing. The suggestion engine filters elements by applicable_entry_points. A filing created against SIMPLIFIED_XBRL will see a smaller element set than FULL_XBRL — and if the line items genuinely belong to the full statement, the simplified subset has nothing close.
  2. Line-item labels are non-standard. The source DOCX uses internal client wording (e.g. "FY24 Op-Ex Pool") instead of market-standard wording ("General and Administrative Expenses").
  3. Synonym library is empty for this client's vocabulary. Phase 1 ships a baseline synonym library; clients with idiosyncratic wording may not match.
  4. The line items belong to a different filing variant — e.g. you uploaded SOCE-only line items into a FULL_XBRL filing.

Diagnose ​

sql
SELECT confidence, count(*) FROM xbrl_tag_mappings
  WHERE xbrl_filing_id = '<filingId>' GROUP BY confidence;

If > 80% LOW, the input is wrong. If a small minority LOW, those are individual judgement calls.

Fix ​

  • Wrong entry point: verify the filing's entryPoint matches the source's intent. The entry point is per-filing; if wrong, create a new filing with the correct entry point.
  • Non-standard labels: map manually using the element search. Tick Save as synonym (reviewer / senior only) for labels that will recur in other clients.
  • Genuine vocabulary gap: save synonyms aggressively for the firm's portfolio of clients. The library improves with every filing.

Internal use only — BreezyCorp