Entity Types Reference | TurnkeyGTM Support

Entity types

Upstream extracts 10 types of entities from your documents. Each type captures a specific kind of GTM intelligence.

#	Entity Type	Description	Metadata Fields
1	Pain Point	Customer problems or challenges your team encounters	severity, frequency
2	Objection	Sales objections raised by prospects during the buying process	frequency, stage
3	Feature Request	Product feature asks from customers, prospects, or internal teams	priority, frequency
4	Content Idea	Ideas for content creation surfaced from conversations and feedback	—
5	Competitor Mention	References to competitors found in your documents	sentiment
6	Deal Context	Sales deal-specific information tied to particular opportunities	funnel_stage
7	Market Insight	Industry or market observations that inform strategy	—
8	Messaging Theme	Recurring communication themes across your documents	—
9	Customer Feedback	Direct customer input, quotes, and reactions	sentiment
10	Process Gap	Identified gaps in workflows or operational processes	—

Relationship types

Entities don't exist in isolation. Upstream identifies 8 types of relationships between entities to build your knowledge graph.

Relationship	Description	Example
related_to	A general connection between two entities	A Pain Point is related_to a Feature Request
supports	One entity reinforces or validates another	Customer Feedback supports a Pain Point
contradicts	One entity conflicts with another	A Market Insight contradicts an Objection
caused_by	One entity is the result of another	A Process Gap is caused_by a Pain Point
leads_to	One entity drives or results in another	A Pain Point leads_to a Feature Request
addresses	One entity solves or responds to another	A Feature Request addresses a Pain Point
competes_with	Two entities are in competition	A Competitor Mention competes_with a Messaging Theme
part_of	One entity is a component of another	A Content Idea is part_of a Messaging Theme

Confidence scores

Every extracted entity receives a confidence score from 0 to 100. This score reflects how certain Upstream is about the extraction.

Score Range	Meaning
80–100	High stored confidence, not proof of truth or approval. Document ingestion may auto-approve under its configured policy; artifact-derived suggestions still require review.
50–79	Medium confidence. The entity is likely present but may need your review. These go to your Review Queue.
0–49	Low confidence. The extraction is uncertain. Review carefully before approving.

Confidence scores are based on the clarity and context of the source text. You can improve scores by uploading well-structured documents with clear language.

Entity lifecycle

Each entity moves through a lifecycle as it enters your knowledge graph:

Pending — The entity has been extracted and is waiting for review in your Review Queue.
Approved — You approved the entity, or document ingestion approved it under its configured policy. Artifact-derived suggestions require explicit approval. Approved entities are eligible for knowledge retrieval in chat and artifact generation; approval is not independent fact-checking.
Rejected — You determined the extraction was incorrect or irrelevant. The entity is excluded from your knowledge graph.

You can change an entity's status at any time from the Review Queue or Knowledge Browser.

Source categories

Upstream organizes documents by source category so you can track where your intelligence comes from:

Sales Calls — Transcripts and notes from sales conversations
Support — Customer support tickets, chat logs, and call notes
Product Feedback — Feature requests, surveys, and product reviews
CRM — Data imported from your CRM system
Internal — Internal memos, strategy documents, and meeting notes
Marketing — Marketing content, campaign results, and market research

Still need help?