Why WhatsApp reply-time numbers differ: a worked example

If one tool says a friend takes two minutes to reply and another says almost five, neither number has to be a mistake. One may show the median and the other the arithmetic mean. The time window and the definition of a reply matter too.

Here is a small example you can reproduce in the free chat analyzer. We made up every name and message. It is a demonstration of the calculation, not a study of real conversations or a benchmark of people’s relationships.

Try it without sharing a real chat

Open the fictional example file and save it to your device, or save the eight lines below as a plain-text file named sample-chat.txt. Open the analyzer, select Choose your chat export, and choose that file. The results appear automatically. You do not need an account, a messaging login or the Mac app. The browser tool processes the file on your device. After the page has loaded, you can disconnect from the internet before choosing it.

31/10/2026, 09:00 - Alex: Coffee on Saturday?
31/10/2026, 09:02 - Sam: Sounds good.
31/10/2026, 09:05 - Alex: How about ten?
31/10/2026, 09:06 - Alex: The place near the park.
31/10/2026, 09:16 - Sam: Ten works.
31/10/2026, 22:16 - Alex: I booked a table.
31/10/2026, 22:18 - Sam: Thanks!
01/11/2026, 22:18 - Alex: See you next week.

These dates use day/month/year. On 2 November 2026, we ran this sample through the same analyzer script served by the live website. It returned eight messages and three conversation starts:

Person	Messages	Starts	Qualifying replies	Typical reply shown
Alex	5	3	2	6 h 32 min
Sam	3	0	3	2 min

The exact median for Alex is six hours, 31 minutes and 30 seconds. The interface rounds to the nearest minute. The small sample makes each calculation easy to inspect; it is far too small to judge someone’s usual habits.

Which gaps count as replies?

The analyzer sorts messages by their timestamps. A message qualifies for reply-time counting when its sender differs from the immediately preceding sender and the gap is shorter than 24 hours.

Sam’s three reply gaps are 2, 10 and 2 minutes. The ten-minute gap starts at Alex’s last message at 09:06, not the earlier message at 09:05. Consecutive messages from one sender do not create extra replies for that sender.

Alex’s two qualifying gaps are 3 minutes and 13 hours. The next day’s message arrives exactly 24 hours after Sam’s previous message, so it is excluded from reply-time counting. It still counts as a message and as a new conversation start.

This is a timestamp rule. It does not inspect whether a message actually answers a question. In a group, a change from one sender to another can be counted even if the second person is speaking to someone else. Quoted replies and read receipts are not used to match a reply to its intended message.

Median versus mean

Sort Sam’s gaps: 2, 2, 10. The middle value is 2 minutes, the median shown by Tera Chat. The arithmetic mean is (2 + 2 + 10) / 3, or 4 minutes 40 seconds.

Both describe those three gaps, but they answer different questions. The median is the middle observed gap. The mean is the total waiting time divided by the number of qualifying replies. A long pause can pull the mean upward much more than the median.

With an even number of observations, the median averages the middle two values. Alex has just two qualifying replies: 3 and 780 minutes. Their median is 391.5 minutes. In such a small sample, the median can still be heavily affected by one long pause.

Why the reply window matters

The current WhatSync analyzer documentation describes a median that skips gaps over 12 hours. Tera Chat excludes gaps of 24 hours or longer. Those rules would treat Alex’s 13-hour gap differently: it qualifies for Tera Chat, but falls outside the window WhatSync describes.

That is an explanation based on WhatSync’s published rule, not a claim that we ran this sample through its software. When comparing results, check the actual implementation or documentation rather than assuming every tool uses the same definition.

Tera Chat’s 24-hour cutoff does not remove every overnight gap. An answer after eight hours of sleep can qualify. Neither a 12-hour nor a 24-hour window knows when someone was asleep, at work or without their phone.

What counts as starting a conversation?

The first message in the imported file counts as a start. After that, a gap of six hours or more starts another conversation, regardless of who wrote the previous message.

Alex gets three starts in this sample: the opening line, the message after the 13-hour pause, and the next day’s message. Sam gets zero. That does not mean Sam never initiated a conversation outside the eight lines we supplied.

A new start can also be a qualifying reply. The six-hour start threshold and 24-hour reply window are separate rules. The 13-hour message in this example satisfies both.

Use the numbers for the question they can answer

Check the date range and messages in your export before interpreting a result. Missing older messages, selecting only one Instagram JSON file from a split conversation, or comparing different date ranges can change the totals. Two tools may also count media or deleted-message placeholders differently.

These statistics can describe the messages present in a file. They cannot establish when a message was read, why someone replied late, whether they care about you, or who is responsible for keeping a relationship going. A direct conversation is more useful than treating a percentage as a verdict.

For your own file, open the chat analyzer or the focused WhatsApp reply-time checker. For a choice of tools with different calculations and outputs, see our chat analyzer comparison.