Article summary
A field-level walkthrough from raw phones to TXT upload, WS Registration and Excel reconciliation, including format exceptions and permission boundaries.
A trustworthy WhatsApp number-checking result is not defined by a successful upload. It is defined by whether every returned row can be reconciled to the source record. This tutorial answers one narrow question: how to prepare an existing phone list as TXT, run WS Registration in AIPUSH, and validate the phone and registration-result columns in Excel.
Build an internal crosswalk before exporting
Do not upload a copied CRM column immediately. Create a staging table with contact_id, phone_raw, country_hint, source and permission_status. contact_id reconnects the result; phone_raw preserves evidence when normalization needs investigation. Names, orders and addresses remain outside the upload.
| Staging field | Uploaded? | Purpose |
|---|---|---|
| contact_id | No | Internal reconciliation after return |
| phone_raw | No | Preserves the original evidence |
| phone_normalized | Yes | Creates the one-phone-per-line TXT |
| source / permission | No | Controls whether later contact is allowed |
Normalize phones without guessing their country
Remove presentation characters, add a country code only when reliable country evidence exists, and apply the relevant domestic trunk-prefix rule. A number without country evidence should enter an exception queue rather than be assigned a country by length. Store long phones as text in spreadsheets so scientific notation does not alter digits. Produce a duplicate report instead of silently dropping repeated rows.
Create a phones-only TXT file
AIPUSH accepts TXT only. Put one normalized phone on each line without a header, name or note. Before upload, record the file hash, total rows, unique rows, blanks and normalization-rule version. If two runs later disagree, the team can first establish whether their input was actually identical.
Choose WS Registration for a registration question
Use WS Registration when the decision only needs the task’s WhatsApp registration result. Its core export is phone plus registration status. WS Activity, Avatar, Gender and Age, and Full Format answer wider questions. Selecting a broader task for a two-field decision creates needless interpretation and access obligations.
Excel acceptance is more than a registration percentage
| Control | Calculation | Warning signal |
|---|---|---|
| Row count | Unique inputs versus returned records | Unexplained missing or additional rows |
| Reconciliation rate | Rows mapped to contact_id / returned rows | Inconsistent normalization |
| Status completeness | Explicit statuses / returned rows | Blanks converted to “no” |
| Duplicate conflicts | Different states for one phone | Batch, time or import defect |
Reconcile row by row in staging
Load the workbook into a staging domain, append batch_id, checked_at and task_name, then join through the normalized-phone crosswalk. Unmatched records go to exception_queue. One-to-many and many-to-one matches go to conflict_queue. A single run must not overwrite the CRM’s master phone or customer-provided identity data.
Keep four result states distinct
An explicit positive or negative result, a blank, an unknown state and a processing error are different observations. An import script must not coerce every empty string to false. If a phone has different results on different dates, preserve both dated observations before deciding whether the platform state or the input rule changed.
Registration is not marketing permission
WS Registration addresses a platform-state question. It does not establish who controls the phone, whether the person expects a message, whether the relationship is current or whether purchase intent exists. Every downstream queue still passes provenance, channel-specific permission and suppression. An opted-out record remains suppressed even when registered.
What the handover package should contain
A reproducible run retains the source count, TXT hash, rule version, task name, observation time, returned Excel, exception list and accountable owner. That evidence turns WhatsApp number checking from a button click into an explainable data process that another analyst can review and rerun.
