Systems that answer
Conversation
One exchange between a person and the system, from their first message to the point it is over — and where that point is decides every rate computed below it.
Also called Session · Thread · Диалог
The definition is a choice, not a fact, and the choice moves every number. A thirty-minute gap ends it, or an hour, or a day; a returning customer next week starts a new one, or continues the old. Set the window at thirty minutes and one buyer over an afternoon becomes four conversations. Set it at seven days and four become one. The answer rate, the refusal rate and the cost per qualified conversation all divide by this, and none of them can be compared with anybody else's until the window is stated.
There is also the question of what counts as a conversation at all. A message that says only “hi”, a wrong number, a supplier pitching you, a bot testing your form — each arrives in the same channel and each inflates the denominator of every rate that is meant to describe how well the thing serves buyers. Excluding them is defensible; excluding them silently is how a deployment reports an eighty per cent answer rate while answering nothing anybody asked.
So the rule is the same as everywhere else in this glossary: write the definition down once, before the reporting is built, and put it beside the number every time it is shown.
Vendors publish answer rates and containment rates without stating the session window or the exclusions, which makes the figures incomparable by construction — the number can be doubled by changing a timeout. Ask any supplier quoting a benchmark what their window is and what they excluded. The question is rarely answered and the asking is the useful part.
Definitions of the unit. Nothing here is calculated, and almost every argument about a marketing number is really an argument about one of these.
Is an input to
- Answer rate
- The share of conversations the system finished on its own — and on its own is doing more work in that sentence than it looks.
- Refusal rate
- How often the system says it does not know — the one metric on a deployment where zero is the alarming reading.
- Escalation rate
- How often a person takes over — and whether they arrive holding the conversation or starting one.
An error in this figure does not stay in it. It reaches three numbers above it, and each of those is quoted as though it were measured.
The term is a reading taken at a point in a sequence that is already running. These are the sequences, with the place each of them jams.
- From a customer’s question to an answer the business can stand behind
- One question arriving at an automated agent, and everything that has to hold for the reply to be something the company would have said itself. 7 steps, buying the model and treating the material as content work is the arrangement that fails while looking reasonable.
A definition is free. Being answerable for the figure it produces is the part that is bought, and this term is a working part of the engagements below.
- The reply that arrives tonight
- An agent that will not say "I do not know" is not a sales asset. It is a liability with a fast response time.
The definitions are the easy part. Whether the figure on your dashboard was computed this way is a different question, and usually the more expensive one.