Article summary
Use a blind sample, field-family acceptance and stop conditions to decide whether an Eritrea +291 WhatsApp Full Format task should scale.
An Eritrea +291 list should not begin at full scale. Use a provenance-stratified pilot to test formatting, field unknowns and reconnection conflicts, and expand only after pre-set conditions pass.
Cover four row types in the pilot
| Type | Purpose |
|---|---|
| Known valid | Test stable return |
| Format exception | Test quarantine |
| Duplicate phone | Test dedupe and reconnection |
| Expected unknown | Test state semantics |
Write stop conditions before seeing output
Stop for unreconciled counts, clustered source failures, mapping conflicts beyond review capacity or sensitive fields without purpose. Do not lower standards afterward.
Give +291 candidates an evidence level
Never overwrite phone_raw. National notation without country evidence stays outside the pilot-success denominator.
Accept three field families separately
Base/activity, profile and account relationship each have pass conditions. A critical failure in one prevents expansion.
A pilot is not a country accuracy claim
It validates the current provenance and workflow only. It does not represent Eritrea’s population or support universal accuracy.
Separate TXT from the control table
Upload one phone per line. Control type, customer context and permission stay internal, and batch_row reconnects output.
Scale through staged quotas
Add one provenance or one volume step at a time and recheck exceptions so a defect does not expand with the whole list.
Failure is useful output
Keep stop reason, affected source and repair recommendation. Do not delete the failed pilot.
The Eritrea value
Evidence decides whether Full Format is worthwhile—not column count or sunk task cost.
