fact_treble_session_nodes

Each row is one interactive step of one conversation through its flow: a question asked, an AI step, a WhatsApp Flow, an integration call — with the option the user chose or the free text they typed, and whether they timed out instead of answering. This is the table for funnels — it did not exist in the previous warehouse.

Steps that only send a message and move on (no interaction) don’t generate a row here — the full message stream, including those, is in fact_treble_session_messages. The complete structure of every flow, all nodes included, is in dim_poll_nodes.

What questions does it answer?

What is the step-by-step funnel of a flow — how many users reach each node, and where do they drop off?
Which answer options do users choose at each question?
Which questions time out most often?
Which conversations went through AI nodes (join with dim_poll_nodes.node_type)?
How long do users take to answer each question?

Columns

Column	Type	Description
`node_record_id`	Int64	Unique identifier of this node visit
`company_id`	Int32	Your company (filtered automatically)
`session_id`	Int64	The conversation flow this visit belongs to
`poll_id`	Int32	Flow identifier
`poll_name`	String	Flow name
`cellphone`	String	User’s phone number
`treble_id`	String	Contact identity (country code + cellphone)
`node_id`	Int32	The flow node visited (join with `dim_poll_nodes`)
`node_type`	String	Step type: `QUESTION`, `AI_TASK_AUTOMATOR`, `AI_AGENT`, `AI_LEAD_QUALIFICATION`, `AI_OBJECTIVE_AGENT`, `WHATSAPP_FLOW`, `META_EVENTS`, `HELPDESK_INTEGRATION`; empty when the node no longer exists
`hsm_id`	Int32	HSM template sent at this step, if any; `0` otherwise
`answer_option_id`	Int32	The predefined option the user’s reply matched; `0` for open answers
`answer_option_text`	String	Label of that predefined option
`response_text`	String	The user’s actual reply text
`is_timeout`	Bool	`true` if the user never answered and the step timed out
`entered_at`	DateTime64	When this step was sent to the user
`responded_at`	DateTime64	When the user answered; `NULL` if they didn’t
`question_msg_id`	Int64	The message that carried the question (`fact_treble_session_messages.message_id`)
`synced_at`	DateTime64	When this row was last written/corrected

Example queries

Funnel of a flow — reach per node

SELECT
    node_id,
    any(node_type)            AS node_type,
    uniqExact(session_id)     AS conversations_reached,
    countIf(is_timeout)       AS timeouts
FROM fact_treble_session_nodes
WHERE poll_id = {your_poll_id}
  AND entered_at >= today() - 30
GROUP BY node_id
ORDER BY conversations_reached DESC

Answer distribution of one question

SELECT answer_option_text, count() AS answers
FROM fact_treble_session_nodes
WHERE node_id = {your_node_id}
  AND entered_at >= today() - 30
  AND responded_at IS NOT NULL
GROUP BY answer_option_text
ORDER BY answers DESC

Conversations that used AI

SELECT uniqExact(session_id) AS ai_conversations
FROM fact_treble_session_nodes
WHERE node_type LIKE 'AI_%'
  AND entered_at >= today() - 7

Get Started

Meta

Build with Treble

Main Platform

Agent Platform

Integrations

Analytics Warehouse

Data Warehouse (Legacy)

fact_treble_session_nodes

fact_treble_session_nodes

What questions does it answer?

Columns

Example queries

Funnel of a flow — reach per node

Answer distribution of one question

Conversations that used AI

​fact_treble_session_nodes

​What questions does it answer?

​Columns

​Example queries

​Funnel of a flow — reach per node

​Answer distribution of one question

​Conversations that used AI

fact_treble_session_nodes

What questions does it answer?

Columns

Example queries

Funnel of a flow — reach per node

Answer distribution of one question

Conversations that used AI