Analytics and funnels

Objective. Use the metrics Fluenta records to identify where conversations stop and make one targeted flow improvement.

Prerequisites

A published bot with conversation activity in the period you want to inspect.
A specific customer journey or flow you want to improve.

Choose the right range and read the overview

1Open Analytics & funnels and select the date range that matches the question you are investigating. A conversation belongs to that cohort when it started in the selected range.
2Read Conversations for the conversation count and completion rate. Completion is the share of those conversations whose outcome is completed.
3Read Order conversion as the share of cohort conversations with at least one order, alongside the total order count.
4Read Handoff rate as distinct conversations with tickets opened divided by the cohort conversation count. The overview also separates ticket reasons into node, keyword, and manual escalations.
5Read KB deflection as conversations answered by knowledge divided by conversations that asked knowledge. It also shows the recorded knowledge-answer invocation count.

Read a flow funnel node by node

1Choose the flow in the funnel selector. The funnel shows how many distinct cohort conversations reached each node and how many were abandoned there.
2Treat an abandonment as a signal, not a diagnosis: it means the conversation’s last turn was waiting at that node and the conversation did not complete.
3Open a node to inspect its turn exits. The drill-down separates turns that proceeded, were reprompted at the same node, completed, or abandoned, and it lists the most-travelled next hops. A conversation that revisits a node can contribute more than one exit.

Turn a drop-off into one focused change

1Prioritize a node with both meaningful reach and a high abandonment rate. A small sample can make a rate look dramatic without representing a broad problem.
2Use the node exit breakdown to form a hypothesis: repeated prompts can point to an unclear question or validation loop; a common next hop can reveal where the current branch sends people; abandonment can justify clearer copy, simpler choices, or a better fallback.
3Change that node or its connected path, test the revised flow, publish a new version, then compare a later date range. The funnel describes observed behavior; it does not prove why a customer left.

Expected result. You can name the flow node where conversations most often stop, explain what its exit data shows, and test one specific change to that path.

Troubleshooting

The funnel is empty: Confirm that the bot has a published flow, select the correct flow, and widen the date range until it includes conversations that reached that flow.
The funnel total does not match the overview conversation count: The overview counts the bot’s conversation cohort, while a funnel counts only cohort conversations that entered the selected flow. A conversation can exist without reaching that particular flow.
A node has a high abandonment rate but the next action is unclear: Open the node detail and inspect reprompts, next hops, completion, and abandonment before changing the flow. Use the pattern to choose a focused experiment rather than treating the rate as a cause by itself.
KB deflection is zero or unavailable: Check whether conversations in the selected cohort actually reached a Knowledge node. Deflection needs a knowledge question and at least one answered result; when no conversations asked knowledge, its rate has no nonzero denominator.