Measurement decision
Telegram attribution should preserve a chain of evidence rather than force one last-click source. Record every relevant community, the source used for selection, campaign and account events, the reply and qualification path, and the commercial outcome. Report both first discovery and campaign influence, then use confidence labels where identity or causality remains uncertain.
A person can belong to five Telegram groups, see your brand in a sixth channel, receive a message from one account, speak to another rep, and convert after an offline introduction. Calling that deal Group A revenue may be convenient. It is not an honest description of the path.
Useful attribution does not need to solve causality perfectly. It needs to preserve enough evidence for better decisions. This guide defines the events and fields that make that possible, the reports that remain defensible, and the claims you should refuse to make.
What the model measures
01An event model from discovery to commercial outcome.
02Separate first-source, selection-source, and influenced-source fields.
03Identity and confidence rules for overlapping communities.
04Decision-ready reports that avoid false precision.
Start with three kinds of source
Never overwrite first observed source when a lead appears again. Add the new observation to a source history. Then record the selection source on the campaign-recipient event. This preserves both discovery history and the reason for a particular contact.
Influence is not proof of causation. Label it accordingly. A defensible report can say that qualified conversations were associated with a set of sources. It should not claim that group membership caused revenue unless the method can support that conclusion.
| Source field | Question it answers | Example |
|---|---|---|
| First observed source | Where did this identity first enter the system? | A developer community |
| Selection source | Which source justified this campaign inclusion? | A partner-specific group |
| Influenced sources | Which known touchpoints may have contributed? | Two groups, a webinar, and a referral |
Use events instead of mutable summary fields
A summary field such as Last campaign is useful for display but weak for analysis because it changes. Events allow the system to reconstruct the path and correct earlier classifications without deleting history.
Use stable internal IDs. Usernames, display names, and group titles can change. Telegram's API documentation also treats authorization and users through identifiers rather than visible labels. Preserve visible values as time-stamped attributes, not identity keys.
| Event | Minimum fields | Why it matters |
|---|---|---|
| lead_observed | lead ID, source ID, timestamp, evidence type | Preserves discovery history |
| lead_qualified | criteria version, decision, reason, reviewer | Explains audience selection |
| message_attempted | campaign, account, template version, timestamp | Separates intent from delivery |
| message_delivered or failed | recipient, state, reason | Prevents inflated activity |
| reply_received | conversation, timestamp, response class | Measures real interaction |
| lead_stage_changed | old stage, new stage, actor, reason | Makes pipeline movement auditable |
| outcome_recorded | outcome type, value range if appropriate, confidence | Connects work to business result |
Resolve identity with confidence, not optimism
A false merge can contaminate contact history and suppression. In an agency, it can also cross client boundaries. The cost of a duplicate record is usually lower than the cost of attaching one person's objection or conversation to another person.
- Exact match: the same stable platform identifier. Merge automatically within the allowed ownership scope.
- Strong candidate: consistent identifier mapping from a trusted import. Review before merge.
- Weak candidate: similar username, name, biography, or organization. Keep separate until verified.
- Conflict: evidence points to different people or client contexts. Do not merge.
Build reports around decisions
Source quality report
Compare reviewed profiles, qualified profiles, useful replies, and qualified conversations by selection source. This reveals whether a group is producing relevant people, not just names.
Handoff report
Measure time from reply to ownership, time to first human response, and stage progression. This identifies value lost after the campaign has already done its job.
Learning report
Group objections, referrals, and disqualification reasons by source and criteria version. Use the report to revise targeting and claims.
Refuse five seductive attribution claims
- A member export represents the reachable or active market.
- The last source observed deserves all credit.
- A reply proves positive intent.
- A deal following a campaign was caused by that campaign.
- A high-volume source is a high-quality source.
Better language: report observed paths, associated outcomes, confidence, and the exact attribution rule used. Precision in method is more credible than precision in a percentage.
Research note
This is an operational attribution model. It does not claim causal identification from observational CRM data. Values, identities, and source evidence should be processed under appropriate privacy, contractual, and ownership rules.
Turn the numbers into a decision
Begin by adding source history and campaign-recipient events. Those two records solve most of the damage caused by overwriting a lead's origin or treating a list as one undifferentiated audience.
The purpose of attribution is not to make every deal look explainable. It is to make the next targeting, staffing, and campaign decision less dependent on memory.
Preserve the path: TeleBoost lead management keeps source lists, campaign contact state, replies, and pipeline context connected so reports can follow the work rather than reconstruct it later.
Continue the operating system