Skip to content

processed_ke_approved

SQL type: SimpleAggregateFunction(anyLast, Nullable(UInt8)) — 0 or 1 Table: keywords.keywords_data_local Customer-facing: No (internal quality flag) Last validated: 2026-05-21

What it is

The pipeline's quality flag. 1 if the keyword passes every quality check; 0 if any check fires. This is the master "do we trust / display anything for this keyword?" signal — every customer-facing derived field is gated on it.

How it's computed

A sequence of AND-combined quality checks in processing.py around L2002–L2025. If any of these fails, processed_ke_approved = 0:

Check Rule Anchor / location
Length keyword text ≤ 100 characters inline check
Word count ≤ 32 words is_gibberish_query family
Not gibberish is_gibberish_query(keyword) returns False processing.py:L210
Not a Google search operator is_google_search_operator(keyword) returns False processing.py:L180
Legit GSC impressions is_legit_impresions(impressions_array, month_array) returns True processing.py:KB-ANCHOR:is-legit-impressions-3mo
Not JS-garbage js_first_seen AND gsc_first_seen aren't both missing inline
Not invalid-long-with-no-data filters keywords that are >50 chars AND have no GSC/JS/GT/GKP signal inline
Not zero-GT-with-high-GSC anomaly filters the specific anomaly pattern where GT is flat-zero but GSC is high (bot trap) inline

Gate semantics

  • = 1: keyword is approved. processed_keyword_volume, processed_volume_trend, processed_growth, etc. are all computed normally and may be displayed (subject to their own gates like volume > 200).
  • = 0: keyword failed approval. Most derived columns are written as null or 0, and the keyword is dropped from PE / customer-facing surfaces.

processed_ke_approved == 1 is a necessary precondition for the volume-200-display-gate — outer_gate_passed = (processed_ke_approved == 1) AND (volume > 200 OR js > 100).

Edge cases

  • Re-approval after data arrives — a keyword that was previously unapproved (no data) can flip to approved when GSC data lands. The pipeline writes the new processed_* columns on the next iteration; anyLast semantics on the table mean the old nulls are overwritten.
  • Re-disapproval is rare — once approved, keywords typically stay approved unless their data goes stale beyond the freshness thresholds.

Downstream

  • Gates every customer-facing field.
  • Read by pe_update.py for PE-tracking selection.
  • Indirectly drives the row-inclusion query: the upload-step query includes WHERE … OR processed_ke_approved = 1.

See also