Average reply time, and why WholeThing uses the median

Updated September 30, 2026

WholeThing’s reply time isn’t an average. It’s the median: the middle of each person’s replies, so one answer that took all day can’t skew it. It measures the first reply in each conversation, the one answering someone who reached out, and points out when a person’s typical reply time changed by 1.8 times or more.

What counts as a reply

Before timing anything, WholeThing decides what a reply is. A reply is the first message you send after the other person’s messages, inside the same conversation. If Sam sends four messages in a row, then you answer with three, that’s one reply from you, timed from Sam’s last message to your first.

“Inside the same conversation” matters. WholeThing starts a new conversation after six hours with no messages, as the who texts first page explains. A message after that isn’t counted as a slow reply. It’s counted as starting a new conversation. So every reply time WholeThing measures is under six hours, and an answer that came the next morning shows up on the conversation-starts card instead.

That’s a deliberate choice. Mixing “answered in four minutes” and “picked it back up two days later” into one number would describe neither.

Why it times the first reply of each conversation

Most replies in a chat are fast for a boring reason: both people are already looking at the screen. In the real chats WholeThing was tested on, about three quarters of all replies arrived within a minute. A typical reply time taken over everything would mostly measure typing speed.

So WholeThing times the reply that matters to most people: the first reply in each conversation, the one that answers someone who reached out. That’s the number a person means by “how long do they take to get back to me”.

If each of you has at least 25 of those first replies, the card uses them.
If not, it falls back to every reply, and then needs at least 40 from each of you.
Below that, there’s no reply-time card at all.

One detail from the file itself: some WhatsApp exports carry no seconds, only minutes. There, a reply that came within the same minute can’t be measured, so it’s recorded as one minute rather than as an impossibly fast zero.

Median, not average

People search for “average reply time”, but an average is the wrong tool here. Take nine replies of about ten minutes and one that took five hours: the average says 39 minutes, which describes none of them. The median, the middle value once the replies are lined up from fastest to slowest, says about ten minutes, which is what it was actually like.

WholeThing uses the median for every reply time on the card. On the card it’s written the way you’d say it: seconds under a minute and a half, minutes under an hour and a half, then hours and minutes, like “18 min” or “1h 34m”.

How it spots a change

A reply time on its own is rarely the interesting part. The change is. WholeThing cuts the chat into periods sized to how long it runs: months for a chat under six months, quarters for one under two years, half-years beyond that. Then, for each of you, it works out the typical reply time in each period.

It compares each period with the one before it, not the first with the last, so a slow drift over years doesn’t read as a sudden break. A change is reported only when:

both periods have enough replies to trust, at least 12 first replies each (or 30 replies each, if it’s using all of them), and
the typical reply time got at least 1.8 times slower or faster from one period to the next.

If several changes qualify, for either of you, the biggest one wins. The card then says who changed and when: “Around July 2025, Sam started taking longer to reply,” with the two periods side by side, here 18 minutes before and 1 hour 34 minutes after. If nothing changed that much, the card shows each person’s typical reply time instead.

Reply rhythm

Around July 2025, Sam started taking longer to reply.

Apr – Jun 2025

Jul – Sep 2025

Sam’s typical reply time, month by month, in 2025.

Sample report. Names and numbers are illustrative.

For a change in the whole chat rather than one person’s replies, see the week it changed, which looks for the single week the conversation shifted most.

Which chats it works for, and when it’s free

Reply times need timestamps, so this card needs a full WhatsApp export. Screenshots don’t show when each message was sent, and WholeThing won’t invent timing from the order of the bubbles. The export guide walks through getting the whole chat out on an iPhone.

The first two findings in every report are free: the strongest one, and the one that answers the question you picked. Reply rhythm is the direct answer to “Is this changing over time?” for a crush and “Has our rhythm changed?” for a partner. It backs up questions about balance and about when things shifted, and it’s usually the opening card when you pick “Show me everything” for a situationship. Everything else unlocks for one conversation, or with a 7-day pass for every conversation you start analyzing that week.

What it can tell you, and what it can’t

It can tell you how long each of you typically takes to answer when the other reaches out, and whether that moved, and roughly when. That’s hard to see by scrolling, because memory keeps the slow replies and forgets the fast ones.

It can’t tell you why. WholeThing sees when an answer was sent, not when a message was read. It doesn’t see calls, a new job, a different time zone or a phone left in another room. A slower reply time is a count, not a verdict on anyone’s feelings.

When a report is unlocked, this card ends with one general line about chats, never about the two of you: “Reply speed usually follows how busy someone’s days are, and it often shifts with work or sleep.”

Is there a normal reply time?

Not one that research agrees on, and the studies that exist are small. Two from the same research group give a sense of scale:

In a 2014 study of 24 people’s phones over two weeks, the median time to see a new message was 6.15 minutes. Asked how fast they usually reply, half said within a few minutes and half within an hour. (Pielot and colleagues, CHI 2014.)
A 2015 follow-up logged 55,824 messages from 42 people. When people weren’t paying attention to their phone, in 75% of cases they were back within five minutes. (Dingler and Pielot, MobileHCI 2015.)

Both measured when a message was seen, not answered, so they’re not reply times. And research on how fast people respond in spoken conversation, where quick answers go with feeling connected, works in fractions of a second, not minutes (Templeton and colleagues, PNAS 2022). That’s why WholeThing compares a person with themselves over time, rather than with a norm that doesn’t exist. How it works covers the rest of the report.

Questions

Why is the reply time shorter than it feels?: Answers that came after six or more hours of quiet aren’t counted as replies. They start a new conversation, so they appear on the conversation-starts card instead.
Is it an average?: No, it’s the median: the middle reply once they’re lined up from fastest to slowest. One reply that took hours can’t drag it up.
Can I get reply times from screenshots?: No. Screenshots don’t show when each message was sent. Reply rhythm needs a full WhatsApp export.
Does a slower reply time mean they’ve lost interest?: WholeThing doesn’t say that. It shows when someone’s typical reply time changed, and work, sleep or a new routine can all sit behind that.