Article summary
Investigate blank avatars in Guatemala +502 lists by partner and collection batch, separating format, age, visibility and task unknowns.
A high blank-avatar rate in a Guatemala +502 partner list does not prove that the contacts have “poor profiles.” The issue may cluster in one acquisition year, format transformation or reconnection path. Audit the partner before judging the result.
Create a provenance comparison
| Source | Comparison | Signal |
|---|---|---|
| First-party forms | Baseline | Current rules |
| Partner A/B | Report separately | Collection difference |
| Legacy CRM | By year | Freshness decay |
Align the phone denominator first
Preserve phone_raw and require country evidence for +502 candidates. Remove format exceptions from the avatar denominator while keeping them visible in quality reporting.
Split a blank into four states
not_observed, visibility_limited, task_unknown and format_exception cannot become one no_avatar value or a fake-account label.
The task field boundary
Profile Picture may return phone, business-account observation, avatar and mapped phone. Business status is not certification, and mapping requires human confirmation.
Examine decay by collection time
If blanks concentrate in older batches, freshness may affect observation. Recheck only a small cohort with a current purpose, not every legacy list periodically.
Partner acceptance is more than return rate
Score provenance completeness, format exception, unknown, conflict and deletion execution. High avatar return cannot offset missing traceability.
Do not judge customers from avatars
Never infer ethnicity, religion, wealth, occupation, personality or purchase power. Avatar supports only a bounded visibility question.
Compare inside staging
TXT contains one phone per line and Excel reconnects through batch_row. Do not publish avatar state to ordinary sales views before the investigation closes.
The Guatemala conclusion
Locate which process creates the blank before deciding to recheck. Blaming contacts for a source defect damages both data and operations.
