Your reply time is two different numbers, and they disagree on purpose

Myth: a chat has one average reply time

It does not. "Reply time" is a label people put on at least two different measurements, and the measurements answer different questions.

One question is: when somebody opens a conversation, how long does that message sit there before anyone responds? The other is: once two people are already mid-exchange, how fast do messages land on top of each other?

A chat of roughly 29,000 messages can answer the first question with a figure measured in hours and the second with a figure in the low tens of seconds. Neither number is wrong. Neither is a rounding error of the other. They are measuring two different human behaviours that happen to share a name in casual speech.

This is the single biggest reason two tools handed the same export can print reply times that look irreconcilable. Before you decide one of them is broken, find out which question each one is answering.

Reality: the waiting number and the pace number

The waiting number, which we call reply wait, starts at the first message of a conversation and stops at the first message from the other person. That is the interval that matches how the wait actually feels. You sent something into silence, and the clock ran until the silence broke. Across a whole year of chat, we take the median of those waits rather than the mean, because a single message answered four days later would otherwise drag the headline figure somewhere no ordinary day resembles.

The pace number, which we call burst reply time, ignores openings entirely. It looks at the gap between consecutive messages inside a conversation that is already running, and it lands in the seconds range because that is not a measure of responsiveness at all. It is a measure of typing speed, thumb habits, and how often a thought gets split across three messages instead of one.

Put plainly: the waiting number is about attention, the pace number is about rhythm. If you want to know whether someone leaves you on read, read the waiting number. If you want to know what your chat sounds like when it is alive, read the pace number.

Myth: the 30-minute silence rule is arbitrary trivia

The session boundary is the load-bearing part of the whole calculation, not a footnote. A conversation ends after 30 minutes of silence. Everything after that silence is a new conversation with its own opening message.

Remove that rule and the pace number stops meaning anything. Your last message at 11:40pm and their first at 8:15am would register as a single nine-hour gap between two messages in an ongoing exchange, which is an absurd description of what happened. Nobody spent the night composing a reply. They went to sleep.

With the rule in place, that overnight gap is not a slow reply inside a conversation. It is the end of one conversation and the start of another, and it gets counted where it belongs: as an opening wait in the morning session. That is also why overnight replies and long silences are worth looking at as their own patterns rather than as noise inside an average.

So when a tool reports a reply time of several hours with no session logic behind it, what you are usually seeing is sleep, work, and travel averaged into a statistic about texting.

Myth: a message nobody answered is an infinitely slow reply

Some conversations never get a reply. Somebody opened, nobody came back, and the next thing in the file is a new opening two days later.

There is no honest number for that wait. It did not take 48 hours, because the reply never arrived. Treating it as 48 hours would quietly invent a reply that does not exist and inflate every waiting figure it touches. So conversations that were never answered are excluded from the reply wait calculation rather than assigned a fake duration.

This matters when you compare tools, because the choice is invisible in the output and it moves the number a lot. A tool that counts unanswered openings as very long waits will always look slower than one that excludes them, on identical data. And if the thing you actually want to know is how often messages go unanswered, that is a separate question with its own answer, closer to ghosting and conversation-starter patterns than to reply speed.

Myth: median and average are close enough

In reply-time data they are not, and the reason is structural rather than statistical hair-splitting. Waits have a hard floor of zero and no ceiling. Most are short, a handful are enormous, and the mean gets pulled toward the handful.

One holiday, one lost phone, one argument nobody wanted to reopen, and a mean reply time drifts hours away from anything either person experienced on a normal Tuesday. The median stays where the mass of the data is, which is why both of our reply figures are medians.

If a tool does not say which one it reports, you cannot compare its output to anything. That is not a small omission. It is the difference between "typical" and "typical plus the worst week of the year".

Reality check: four questions before you compare two numbers

Does it measure openings or in-conversation gaps? A figure in seconds is almost certainly measuring pace. A figure in hours is almost certainly measuring waiting. If a tool reports one number in the middle, ask what it blended.

Does it have a session boundary, and how long is it? Without one, overnight gaps become reply times. With a very long one, separate conversations merge and openings disappear.

What happens to unanswered messages? Excluded, or counted as the gap until the next unrelated message? This alone can double a reported wait.

Median or mean, and over what window? The reply time of a full year and the reply time of the last 90 days are different facts about the same relationship, and so are the median and the mean of the same set of gaps.

Once you know those four answers for both tools, the disagreement usually stops being a mystery and becomes a translation problem. Two clocks, two starting guns, two questions. Read them as a pair and you learn something neither one says alone: how long you wait to be noticed, and how fast things move once you are.