Conversations

People rarely ask an AI engine once. They ask a question, then narrow it down: "best CRM?", then "for a small team?", then "which is cheapest?". A brand that is named in the first answer often disappears by the last. Conversation measurement asks a chain of two or three questions as one conversation on every engine that supports it, and shows how often each brand is still named at every step.

#Where to find it

Prompts › Your Prompts, then the Conversations tab. The band at the top shows how many conversations your plan allows ("2 of 3 conversations") and which engines are measured in this workspace, each marked Supported or Not supported for conversations.

#Which engines can hold a conversation

A follow-up only means something if it is asked in the context of the earlier questions. Some engines answer one question at a time, so a follow-up cannot be asked in context. Those engines are still listed — marked Not supported for conversations — and still measured for your single prompts, but not for conversations. If none of the engines measured in your workspace can hold a conversation, the tab says so and nothing is measured there for now.

#Adding a conversation

Only owners and admins can add, edit, pause or remove conversations.

1. Click Add conversation.
2. Optionally give it a Name; otherwise the first question is used as its name.
3. Write Step 1 and Step 2 — for example "What is the best CRM?" and "Which one suits a small team?".
4. Optionally click Add a third step ("Which of those is the cheapest?"). A conversation has two or three steps, each up to 400 characters.
5. Click Save.

The conversation is asked from the next nightly measurement onwards and then daily, like your prompts. Until the first answers have been read, it shows Not measured yet.

#Reading the results

Select a conversation in the list on the left. Its detail shows when it was Last measured and two views for the chosen period (7, 30 or 90 days).

#How your brand holds up across the conversation

A funnel with one column per step and the same brands in every column: your brand and the three rivals named most often. Each bar shows how often that brand was named at that step, with its confidence interval underneath ("likely between 22% and 41%").

Below the funnel is the Drop-off line: "you were named at the first step but no longer at the last in 7 of 20 conversations". It is calculated only from conversations that named you at the first step and reached the last step.

#Your brand, by engine

A table with one row per engine and one column per step, plus a Drop-off column. It shows where the conversation loses you: one engine may keep you to the end while another drops you after the first follow-up. Engines that cannot hold a conversation stay in the table with the Not supported for conversations badge instead of numbers.

#Not enough data

Like every rate in Rixe, a step's rate needs at least 4 answers before it is shown as a percentage. Until then the cell reads "Not enough data (2 of 4 answers)". Drop-off likewise waits until at least 4 conversations named you at the first step.

#Missing data

If a step could not be measured — the engine failed, or an earlier step of the same conversation failed so the follow-up could not be asked — the detail shows a badge such as "2 steps could not be measured". Answers that are still being read show "3 answers are still being read". Missing steps are never counted as "not named".

#Editing, pausing and removing

• Edit changes the name or the questions. Changing the questions starts the results over: answers to the old wording are no longer counted.
• Pause stops measuring the conversation and keeps its results; Resume starts it again. A paused conversation is marked Paused in the list.
• Remove stops measuring it for good, after a confirmation.

#Plan availability

The limit is counted across your whole account, not per workspace. A conversation costs more to measure than a single prompt, because every step is asked with the earlier turns included; on agency plans each conversation measurement draws credits for every step. When you reach the limit, remove a conversation or move to a larger plan to add another. Conversations already tracked keep being measured after a downgrade. See Plans and limits.

#Related pages

• Prompts — the single questions Rixe tracks.
• How measurement works — the daily cycle and missing data.
• Credits — how agency credits are drawn.