Track Telegram member activity without mistaking volume for value

Imagine one member posting forty messages while another posts one answer that resolves a difficult problem. The first is more active by message count. That is all the count establishes.

Tracking Telegram participation is useful when it helps you make a concrete decision: find a gap in moderator coverage, notice that one volunteer carries the conversation, or check whether newcomers can join in. It becomes misleading when the same numbers are used to label people valuable, lazy or suspicious.

This guide explains how to read group and individual activity in Defendy, check the limits of the data, and turn the result into a proportionate next step.

Start with a question you can act on

Do not begin by exporting everyone into a league table. Choose the decision first and inspect only the information needed for it.

Match the participation question to evidence and its limits
Your question	Useful evidence	What to avoid concluding
When would another moderator help?	Hourly activity plus examples of unanswered or difficult conversations	Everyone posting at that time is available to moderate
Are we depending on a few helpers?	Who contributes regularly, what they do, and whether someone can cover them	The person with the most messages does the most useful work
Are newcomers finding their way?	Arrival information where covered, first questions and the replies they receive	Joining or posting once proves successful onboarding
Did a rule change suppress ordinary discussion?	Comparable periods, moderation events and a sample of affected messages	A fall in messages alone proves the rule was harmful
Should an account be removed?	Specific policy violations and human review	Silence, a missing avatar or a low score proves it is a bot

CHAOSS, a community-health project, treats dependence on a small set of contributors as a continuity question. Its Contributor Absence Factor is designed for open-source work, not as a Telegram member rating. The useful question to borrow is “what work would stop if this person took a break?”, rather than “who should post more?”. CHAOSS's metric description.

Find the right view in Defendy

Open Settings in your private chat with @defendy_bot, choose the group in My groups, and open Statistics. Confirm the group name before interpreting any chart.

Choose 7, 30 or 90 days, or a custom range. Keep the exact dates and the group's timezone in your review note. The dashboard aggregates dates in the group's timezone; the reviewer travelling with a different phone timezone does not make that a different community day.

Overview: the shape of the conversation

The overview contains message count, active authors, average daily activity, weekly activity (WAU), messages per active author and moderation actions, plus daily and hourly charts. WAU counts distinct authors in the last seven local days of the selected range, or the whole range if it is shorter. It is not an average of weekly values.

Here, “active” is tied to recorded message activity. It is not a count of people reading, learning or receiving value. Average daily active authors is an average across days; it is not interchangeable with the number of distinct authors over the full period. One person speaking on several days can contribute to several daily counts but remains one distinct author for the period.

There is another reconciliation trap: the checked overview can include messages sent as channels, while the Users activity table focuses on user senders. Do not expect the sum of individual people in a top list to explain every overview value. The table is also only a top list, not a complete roster of all members.

Users: participation and membership are separate

Open Users for membership changes and the most active users. The activity table includes message counts and active days. A person who posted during a single event and one who answered throughout the week can have similar totals but different participation patterns.

Read the collection-coverage notice. Some membership periods may be only partly covered or unavailable. A missing value is not evidence that nobody joined or left. A total marked as estimated should not be treated as an exact historical headcount. The coverage notice concerns membership data; it does not certify that every message or moderation event has been captured.

More → User analysis: inspect a specific question

For a closer look, open More → User analysis. Select up to five known users and choose the period. The grouping controls offer day, week, month or total. Categories cover messages, membership, moderation and reputation, with the relevant metric selected within the category.

For example, select a consenting volunteer to check whether their contribution is spread across the month or concentrated around an event. Use message counts and active days as context for a conversation about workload. The view does not establish how much time they spent helping, whether an answer was correct or why they were absent.

If you cannot find someone in the picker, verify the group and account identity. Do not treat absence from search as a complete history of non-participation. The interface works with users known to the service, not an independent census of everything a person has ever done.

The statistics reference covers the wider dashboard. For community-level moderation measures rather than individual participation, see the metrics guide.

A calculation that looks simple but needs a definition

The following numbers are hypothetical, chosen only to illustrate interpretation. Suppose your consistently defined review set contains 600 recorded messages from 20 distinct posting members in a week. The average is 600 divided by 20, or 30 messages per posting member.

That does not mean a typical member wrote 30 messages. One person might have written 300. Nor does it mean twenty people were helped. The calculation describes posting volume within that review set, not satisfaction, reading or intent.

Suppose the next comparable week contains 400 messages from 25 posting members. The average falls to 16, but more people have spoken. Several explanations remain possible: answers became shorter and clearer, a prolific member took a break, an argument ended, spam was removed earlier, or collection changed. Read a sample before choosing one.

Do not calculate “engagement rate” by casually dividing these posting members by an estimated current member total and comparing it with another tool. The dates, denominator, included sender types and collection coverage must match. State the definition alongside any percentage you publish.

Check the data before judging the people

A chart is the end of a collection process. Telegram sends updates; the service processes relevant events; the reporting view applies its filters and period. Events missing from that path cannot be reconstructed merely by selecting a wider date range.

Telegram's Bot API distinguishes new messages, edits, reactions and membership updates. It also has separate conditions for delivering some update types. Do not interpret a bot report as an unrestricted archive of the entire group. Telegram's update reference.

In particular, do not assume every deleted message appears in the activity total. A message count and a deletion count are built from different event types and need not describe identical sets. A deletion recorded by a filter does not establish that the same message was counted in the activity view. Counts of moderation actions also do not mean the same number of offending people: one incident can involve more than one action.

Before a consequential interpretation, make these checks:

Dates: compare equal-length periods with comparable weekdays. Prefer completed days when judging a change; today's total is still growing.
Context: note events, launches, holidays, outages and rule changes. A live session is not an ordinary Tuesday.
Coverage: inspect membership notices and estimated totals. Mark unknown values as unknown in your notes.
Identity and scope: distinguish user activity, channel senders and a limited top list. Do not merge people merely because their display names match.
Examples: inspect the underlying conversations that would change the decision, including useful messages caught by moderation.

CHAOSS's activity-time guidance describes aggregation in UTC or local time and explicitly flags privacy and data-ethics risks. For a Telegram admin, the practical application is to use a group-level pattern for scheduling and avoid trying to infer a member's location, job or personal routine from timestamps. CHAOSS on activity dates and times.

Run a small collection check when numbers look wrong

Use a test group and a consenting member. Note the current period, starting values and time. Have the member send one ordinary message, then refresh the relevant view after the data has had time to update. Compare the overview, user row and selected-user analysis using the same date range.

This is a suggested diagnostic, not a claim that every view updates instantly. If the new message is absent, first check group selection, date boundaries, timezone and whether the view reports an error. Use its retry control where offered. Do not invent a growth or decline result from a loading state.

If counts still disagree, record the exact views, filters and test timestamp and request investigation. Never generate a burst of fake conversation to make the graph move. If you test a harmless filter deletion, treat the result as evidence about that specific route only, not proof that all deleted messages are counted.

Act on the finding without creating a surveillance habit

If the group relies on one helpful member, thank them for specific work and ask whether they want support. Document the recurring answer or invite a willing co-host. Do not turn their past activity into an obligation to be available.

If newcomers ask questions that go unanswered, improve the welcome information and assign someone to watch that type of question. If ordinary messages disappear after a rule change, inspect the rule and the affected examples. If a quiet member has broken no rule, there may be nothing to fix.

Keep named analysis accessible only to the people who need it for the decision. Explain in community guidance what is measured and why. For a public update, prefer a group-level summary over a ranked list of quiet members. If you download an export, check its columns, destination and access before sharing it; do not copy names, identifiers or message excerpts into an unrelated public report.

A short review note is enough:

Question and possible decision
Group, exact dates and timezone
Observed numbers and coverage limits
Relevant conversation examples, kept private when necessary
Alternative explanations
One action, its owner and a date to review the result

The output should be a better decision about the group. It should not be a permanent personal scorecard for everyone in it.

Sources and verification

Checked 5 October 2026 against current Defendy statistics views, their data definitions and the linked primary sources. No live community dataset was analysed for this article. All numerical examples are hypothetical; the diagnostic is a proposed procedure. Interface availability does not establish complete historical collection or the intent of any participant.