Ghosting and hot & cold: what the dates show

The difference between a stop and a fade

Both end the same way, a thread that stops moving, and they read in opposite ways in the file.

• A hard stop: volume stays normal right up to the last message, then nothing. The curve is flat then drops to zero. Something specific happened, and the date of the last message is the date of that something.
• A fade: volume drops over months, messages get shorter, questions disappear before replies do. No single date stands out. This is the more common case, and the most invisible from the inside.

The second one is measurable long before it ends, and that is the whole point of looking. In one real conversation from the RevealChat corpus, a thread went from 18,645 messages in a year to 794 the next. Nobody saw it coming: every week looked like the one before.

The three traces that come first

1. 01Questions stop before replies doThis is the earliest and most reliable signal. Somebody keeps replying politely but stops asking anything. It can be counted: number of messages containing a question, per month, per person.
2. 02Average length collapsesMessages go from three lines to three words. "ok", "ah right", "haha". The message count can stay flat while the volume of text divides by four.
3. 03Initiation tips to one sideOut of ten restarts after a silence, nine come from the same person. That is the number that hurts most to see written down, and the hardest one to argue with.

Hot & cold: look at the spread, not the average

A hot and cold pattern has a very recognisable signature: the average reply delay stays normal while its spread explodes. Four minutes one day, two days the next, four minutes after that. An average reply time of six hours can describe somebody steady or somebody unpredictable, and those are two very different conversations to live through.

That is why a single number is not enough. The median says what usually happens, the spread says how much you can rely on it. The three definitions of reply time are broken down here.

The double text, and what it actually counts

Writing twice in a row without a reply is the most direct measure of imbalance there is. It says nothing about the other person’s intentions, it says how many times you reached into the void, and over what period.

One number to read next to it: the share of those follow-ups that get an answer. A thread where eight out of ten go unanswered, and the other two get an enthusiastic reply, produces exactly the confusion people describe as hot and cold.

What these numbers do not prove

A silence in a file could be an argument, a house move, a bereavement, a heavy work period or a broken phone. The file says there was a silence and since when, never why. Any reading that jumps to the cause invents half of what it claims.

And an imbalance is not a character flaw. Some people text little and are present in other ways; others text a lot, to everybody. What the numbers give you is knowing what you are talking about before you talk about it: "since March, I open nine conversations out of ten" is a sentence you can put on the table without accusing anyone.

To see these measures on your own conversation, start with the export. Two neighbouring dynamics read with the same tools: the mental load and complaints that keep coming back.