Article summary
An enterprise runbook for repeatable TG number-checking jobs: intake tickets, task-specific folders, controlled operators, Excel reconciliation and incident recovery.
Enterprise Telegram number checking does not require pretending that a file workflow is an API. It requires a runbook that makes every transfer observable: who requested the job, which TG task was approved, what TXT was submitted, which Excel result was received and how exceptions were resolved.
AIPUSH currently uses TXT input and Excel output. The runbook below is designed for teams that operate recurring batches across markets, business units or CRM systems.
The intake ticket controls scope
No file should be prepared until the requester completes a ticket with business purpose, source system, record owner, country scope, authorization evidence, selected TG task, requested completion time and retention period. The data owner approves the purpose; the operator does not.
If the requester writes only “Telegram screening,” return the ticket for a task decision. Registration, activity, username, gender/age and full-format services return different fields.
Use a folder layout that tells the batch story
TG-20260911-018/
00-request/
10-source-private/
20-outbound-txt/
30-inbound-excel/
40-reconciliation/
50-approved-output/
90-audit-log/
Restrict 10-source-private more tightly because it contains internal identifiers and permission context. The outbound folder should contain only the numbers required for the task. No operator should email workbooks between these zones.
Assign roles before assigning files
| Role | Accountability | Must not do alone |
|---|---|---|
| Requester | Defines purpose and source population | Approve their own high-risk expansion |
| Data preparer | Normalizes and exports TXT | Change task after approval |
| Processing operator | Uploads TXT and retrieves Excel | Edit the source extract |
| Reviewer | Checks schema, counts and exceptions | Ignore failed reconciliation |
| CRM owner | Applies approved fields | Activate suppressed records |
Prepare the file according to the ticket
Normalize numbers to a documented international key, place one value per line and remove exact duplicates. Maintain a local bridge from normalized number to source record IDs. Count rejected rows by reason: missing country, invalid length, blank value, duplicate or policy exclusion.
Seal the outbound file after approval. If a row changes, create a new version and record the difference rather than editing the uploaded artifact in place.
Lock the expected TG schema
The registration task returns phone number and registration status. The activity task can return phone number, TG UserID, username, offline time, active days, first and last names, TG VIP state and frozen state. Gender/age and full-format tasks extend the schema with profile-related fields.
Attach the expected headers to the ticket. When the Excel result arrives, the reviewer checks that contract before examining values. An unexpected missing column is a processing incident, not a cue to improvise a new mapping.
Reconcile in two passes
The first pass is mechanical: counts, data types, duplicate keys, normalized joins and missing rows. The second pass is semantic: whether values fit their definitions, unknown states remain unknown, and mapped identifiers are not overwriting source identifiers.
Produce an exception workbook rather than modifying the master result. Each exception receives an owner, reason and disposition. A batch is not complete merely because most rows joined.
Recover from the four common incidents
Wrong task was processed
Quarantine the result, correct the ticket and rerun under a new version. Do not adapt CRM mapping to the accidental schema.
TXT and result cannot be linked
Compare checksums, filenames and operator log. Confirm the workbook belongs to the same request before opening its fields to downstream users.
Phone columns were converted by spreadsheet software
Reimport as text from the untouched workbook. Do not repair scientific notation manually.
Join rate falls below the accepted threshold
Stop application, segment exceptions by country and rule, and correct the normalization code before resubmission.
Release only a minimum approved dataset
The final output for a business team should contain source ID, necessary TG result fields, permission state, suppression outcome and rule version. Remove operational columns the recipient does not need. Broad profile fields should not be distributed simply because they exist in the processing workbook.
Close the job and shorten retention
- Record input, output, joined and exception counts.
- Store reviewer approval and mapping version.
- Apply updates using the batch ID.
- Confirm the activation layer rechecks suppression.
- Delete transient copies on the ticket’s schedule.
- Keep a minimal audit record of actions and outcomes.
Record lessons in the runbook when an exception repeats, so the next team benefits from the repair.
This runbook makes recurring TG checking reproducible without claiming a direct API connection. When each batch has a ticket, artifact chain, schema contract and closure record, the file workflow behaves like a managed enterprise service rather than an informal spreadsheet exchange.
