WhatsApp group statistics explained, number by number

Feed a WhatsApp export into any analyzer and you get a wall of numbers: totals, percentages, peaks, records. Most of them look obvious until you ask a simple question like “does a sticker count as a message?” or “why does my friend in Mexico look like a vampire?”. This guide goes through every statistic Chat Recap shows, explains exactly how it is computed, and points out where the numbers can mislead you. Even if you analyze your chat some other way, the same logic applies.

What counts as a message

Everything starts with one decision: what is a message. In Chat Recap the rule is simple. A message is anything a person sent: text, a photo, a video, a voice note, a sticker, a document, a poll question, even a message they later deleted. What does not count are the notifications WhatsApp generates on its own, such as “Anna added Biel”, “You changed the group icon”, security code changes or missed call notices. Those are set aside as system messages.

Inside that total, messages fall into three types:

• Text messages: anything with written content. Only these feed words, characters, emojis and phrases.
• Media messages: the placeholders WhatsApp writes when you export, like <Media omitted> or image omitted. They count towards each person’s total, but they have no words.
• Deleted messages: lines like “This message was deleted”. They count as something the person sent, but their content is gone.

A couple of clean-up steps also matter. The “This message was edited” tag is removed so it doesn’t pollute word counts. @mentions are stripped from the text before counting words. And when someone reposts a sign-up list (“1. Anna, 2. Biel, 3. Carla…”) each time a new name is added, the repeated lines are only counted the first time, so a long list doesn’t make the same words look hugely popular. If you want the gory details of what an export looks like line by line, see how the WhatsApp export file is structured.

The group summary

These are the headline numbers for the whole chat.

Statistic	How it is calculated
Messages	All messages sent by people (text, media, deleted), without system notices
Daily average	Messages divided by every calendar day from the first to the last message, including silent days
Active days	Days with at least one message, out of the total span
Busiest day	The single date with the most messages
Peak hour, favourite weekday	The hour (0 to 23) and weekday with the most messages overall
Busiest month and year	The calendar month and year with the most messages
Longest streak	Most consecutive calendar days with at least one message
Longest conversation	The biggest burst of messages with no pause longer than 30 minutes
Longest silence	The biggest gap between two messages, only counted if it lasts 3 days or more
Average length	Characters per text message (media and deleted messages are left out)
Words	Total words in text messages, links excluded

Two of these deserve a note. The daily average includes days when nobody wrote, so a group that is very busy on weekends and dead during the week will show a modest average. Compare it with active days to see how “bursty” the group is. And the longest conversation is not about topics: it is purely about timing. Ninety messages sent with never more than half an hour between them count as one conversation, even if they jumped from football to someone’s lost keys.

Voice note minutes only appear when you upload a .zip that includes media, because the duration is read from the audio files themselves. With a “Without media” export, you only get the number of voice notes.

Numbers per person

The participants table ranks people by messages and shows, for each one: share of the group’s messages, words, average length, emojis, links, media, voice notes, night messages (sent between 00:00 and 05:59), favourite hour, favourite weekday, favourite emoji, and the dates of their first and last message.

Here is an invented example of a four-person group with 11,000 messages:

Person	Messages	Share	Avg. length	Words
Anna	4,210	38%	22	17,900
Biel	2,950	27%	61	30,100
Carla	2,400	22%	35	14,000
Dani	1,440	13%	18	4,300

Anna “wins” on messages, but Biel actually wrote far more words. Anna is a machine-gun typist (“yes” / “coming” / “5 min” / “where are you”), while Biel sends paragraphs. Neither number alone tells you who “talks most”. Looking at messages, average length and words together does.

There is also a signature word for each person: a word of at least four letters, used in at least five messages across the group, where one person accounts for 60% or more of the uses. It is the word that, statistically, belongs to them.

When the group talks

The activity section shows messages per month, per year, per hour and per weekday. Months with no messages appear as zero instead of being skipped, so long silences are visible in the chart. You can switch between the last year, the last three years and the full history; the year is counted back from the last message in the chat, not from today.

Two extra charts are easy to overlook:

• Active people per month: how many different people wrote at least once that month. Great for spotting when a group quietly shrank to its core members.
• Participation over time: each month, the share of messages sent by each of the top five people. This is how you see someone take over the group after a breakup, a new job or a new phone.

Words, phrases and emojis

Top words ignore very common words in Catalan, Spanish and English (articles, pronouns, “yeah”, “okay”…), laughs like “hahaha”, words made of a single repeated letter, and any word shorter than four letters. Groups often mix languages, so all three lists are always applied.

Expressions are pairs and trios of words that start and end with a meaningful word, so “the the” never wins but “good morning” can. They need to appear at least twice.

Catchphrases look for short messages repeated word for word (up to seven words) and three to five word expressions that keep coming back inside short messages. Each one has to appear at least three times, and each message counts once.

Emojis are counted individually, including flags and combined emojis. Symbols like © or ‼ are ignored. Note that a thumbs up with a skin tone and a plain thumbs up are counted as different emojis.

Who replies to whom

To estimate interactions, the analysis looks at consecutive messages from two different people sent within 15 minutes of each other, and counts that as a reply. Pairs need at least five of these exchanges to be shown. It is a rough proxy: in a busy group, “replying” to the previous message might just mean writing at the same time.

Traps when reading the numbers

• Media inflates totals. Someone who sends forty memes a day can top the ranking without writing a word. Check the media column.
• Time zones. Times come from the phone that exported the chat, with no conversion. If a member lives seven hours away, their “3 am messages” may be lunchtime for them, and they will look like a night owl.
• Late joiners. Someone who joined last spring can’t compete on totals with founding members. Look at their first message date and their share in recent months instead.
• Small groups. In a chat of three people with 300 messages, one busy weekend decides every record. Percentages swing wildly.
• Exports only go so far. If your export doesn’t reach back to the very first day, everything starts where the file starts.
• Shared phones and renamed contacts. The name shown is whatever the exporting phone had saved. Contacts saved only as a number are shown as “Contact 1”, “Contact 2” and so on.

Try it on your own group

Export the chat (the step-by-step export guide covers Android and iPhone) and drop the file into the analyzer. The statistics are calculated in your browser, without AI. Then check the group awards, which turn several of these numbers into titles, and see who ends up as the ghost.