Metrics and terms
Counting conversations, and the little it proves
Volume is the easiest number in the product to produce and the easiest to over read. It rises when a campaign lands, when the launcher moves somewhere more prominent, and when your site gets more traffic for reasons nobody in support influenced. It is genuinely useful, but almost entirely as a denominator.
shown in the product
The tile is labelled "Conversations" and hinted "last 30 days". It counts conversations started in the window. Nothing on that page attributes them to a page, a campaign or a visitor, and no unique visitor count exists, so a returning visitor appears more than once.
What it means
Conversation volume is the count of distinct conversations started in a window. A conversation is the unit a visitor experiences: one panel, one thread, however many questions they asked inside it. Counting them tells you how many separate people brought a problem, which is a different and more human quantity than the message count sitting next to it. What volume cannot tell you is whether any of those people got what they came for, and it will not tell you whether it saved anybody any work, because a conversation that happened in the widget may also have happened again by email an hour later without anything here knowing.
How it is actually calculated
What counts as one
A conversation created in the last thirty days. The count is of conversations, not of visitors, so somebody who comes back the following week is counted twice and somebody who opens the panel and asks three things is counted once.
The message count next to it is the other half of the picture. Messages divided by conversations gives you the average depth of a conversation, which is a more interesting figure than either on its own and is not displayed for you.
The window and what it hides
Thirty rolling days with no comparison to the previous period on that page. So a figure that has halved looks exactly like a figure that has always been that size.
There is no split by page, by source, or by time of day on that page. If you need to know where conversations came from, that question belongs to your own analytics rather than to this tile.
Why it is a denominator rather than a headline
Almost every other number here becomes interpretable when divided by this one, and almost none of them are worth reading without it. A cache share means something different at forty conversations and at four thousand. An empty unanswered list means nothing at all when volume is tiny.
Treated as a headline it measures the placement of a launcher and the size of your traffic. Both are real things. Neither is evidence the assistant is answering well.
How the number gets moved without anything improving
How the count rises without anybody being helped
Make the launcher larger, brighter, or set it to open on its own. Conversations rise the same day. Every one of the extra conversations is somebody who did not intend to start one, and nothing about your material changed.
The accidental version arrives constantly and from outside your team: a campaign, a link from somewhere busy, a seasonal peak. Volume moves, every ratio on the page shifts with it, and it is easy to read the shift as a change in performance.
This is why the useful reading of volume is as context for other figures rather than as a result on its own. It is the number most likely to be quoted upward in a meeting and the number least likely to mean what the room thinks it means.
What to look at instead, or alongside
- Messages divided by conversations, which shows whether visitors are asking follow ups or leaving after one try.
- The unanswered list, which is volume expressed as work you could do.
- Contacts across your other channels over the same period, if the question behind the question is whether the widget removed any.
- Your site analytics for where conversations start, which this page does not attempt to tell you.
Questions
- Does a visitor who opens the panel and types nothing count?
- The count is of conversations, so what matters is whether a conversation record exists rather than whether anybody was satisfied. When you compare this figure against anything from another product, this is the first definition to check, because vendors draw that line in different places and it moves the number a lot.
- Why is there no unique visitor count?
- Because it would require identifying people across sessions, and the widget deliberately does not do that. The cost is that a returning visitor is counted twice; the benefit is that nothing here builds a profile of anybody.
- Our volume dropped. Should we worry?
- Look at your site traffic first. A drop in widget conversations that matches a drop in page views is a marketing event, not a support one. A drop with flat traffic is worth a look at whether anything changed about where the launcher sits.
Keep reading
- Containment rate: why this product does not compute oneContainment counts conversations that ended without a person. A widget cannot see whether one did, so no containment figure exists here.
- First response time has no meaning without a queueThis metric measures how long somebody waited for a human to pick up. There is no queue in a widget, and no first response time is computed.
- Escalation rate: what a handover is here, and why it is not a rateHandover enquiries are stored and listed. No rate is computed against them, and dividing by conversations by hand needs three caveats.
- Every metric and termWhat each one means, and which of them this product actually computes.
Try it on your own material
Upload a document or point it at your site, paste one line of HTML, then ask it something only your business could answer.