Bring customer evidence into Discovery

For a first test, open Discovery → New → Import data… → Paste text. Import one real call note, ticket, review, or research excerpt and inspect the resulting Signal, extracted Evidence, and related Insights.

This gives you a baseline before you configure a connector or automate a source.

Use manual import when the evidence is already on your computer or you are testing a small batch.

Upload a file

In Discovery, select New → Import data… → Upload file. Choose Upload file for raw customer evidence. Choose Curated findings or Requests & bets only when the file already contains finished records of that type.

Choose the mode that matches your file:

Mode	Use it when	Result
Rows	Each CSV/XLSX row is a separate piece of evidence.	One signal per row; you map columns such as title, date, product, account, and evidence text.
Whole file	The report, export, or document should stay together.	One signal for the file; no column mapping. This is usually the clearest choice for DOCX, PDF, TXT, Markdown, and HTML.

If you are unsure, choose Rows only when each row can stand on its own. Otherwise, choose Whole file.

1. Add your file. You can add multiple transcript files when each file represents one call or meeting.
2. Choose a Signal type—usually Customer feedback for rows or Discovery report for a whole file. Add Product or a Global account when the same context applies to the entire upload.
3. Select Preview. Nothing is created yet. For Rows, check the selected product, account, title, date, evidence, and source-link columns. The preview groups records as:
   • Ready: will become a signal.
   • Review: needs a decision, usually because an account value does not match an existing workspace account.
   • Skipped: will not become a signal.
4. Resolve every account review, then select Import signals. You can map an unmatched value to an existing account, import it without an account, apply the global account, or skip the row. Until the review is resolved, the import action remains blocked.

Zentrik links rows to existing workspace accounts; it does not create accounts during an import. Imported signals retain source details such as the filename, sheet or row, dates, and URLs.

Paste text

Use Paste text for notes, transcripts, tickets, reviews, survey responses, or reports.

Limits

Files can be up to 50 MB each. A batch can include up to 100 files, and a Rows import can create up to 500 signals.

Use a connector when the evidence already lives in a supported system. Connect it under Settings → Integrations, then choose the records to import from Discovery.

Source	Imports	Creates
Zoom, Microsoft Teams, Gong, Fireflies, Granola, Fathom	Calls, meeting transcripts, or notes	Signals
Zendesk	Support tickets and customer issues	Signals
Pendo	NPS responses, comments, and account context	NPS records; optional signals from useful comments
Jira, Linear, Aha!	Issues, requests, and product ideas	Signals, insights, or ideas, depending on the source and mapping
Salesforce	Cases, opportunities, and accounts	Signals, insights, or ideas
Airtable	Requirements, roadmaps, and tracked records	Signals or ideas

Other evidence paths are available through Slack thread capture and workspace email intake. Each source guide explains its permissions, filters, and mapping rules.

Use a one-time import when you want to choose particular records now. Use a saved import when Zentrik should reuse the same source boundary on demand or on a schedule.

1. Connect the source under Settings → Integrations.
2. Open Settings → Integrations → Imports, select New saved import, and define what can be imported and where it belongs.
3. Select Create draft. A source preview is an optional spot check while you refine the rules.
4. Select Test draft. Testing does not import records or change what the active import has already processed.
5. After a successful test, select Activate tested draft. Keep it On demand or choose an interval, then review results under Runs or Imports → History.

Changes to an active import stay in a new draft. The active version keeps running until the changed draft is tested and activated.

Use the API when Zentrik does not have a connector for your source or you control the import workflow yourself.

• Signals API quickstart: send one record and confirm the contract.
• Signals import template: batch records with retries, polling, dry runs, and source mapping.
• Signals API reference: inspect the request and response fields.

For API imports, keep a stable externalId, source date, product or account identifiers, and a reviewer-facing sourceLinks URL when available.

If a source is not listed and you prefer a native connection, contact us. If you already control its API or export, the technical guides can get it into the same Discovery pipeline.

For Signal imports, Zentrik stores the source text as a Signal, processes it into source-backed Evidence, and connects that Evidence to related entities, including Insights. Insight and Idea imports create the chosen record directly. Pendo NPS history remains account context; useful comments can become Signals. Source links and imported metadata stay with the record when the source provides them. Longer runs can continue in the background after the dialog closes.

For raw input, the same Signal → Evidence → Insight path is used for pasted text, files, connectors, and API imports. Results vary with the source: a rich customer conversation may produce several Evidence items; a short note may produce one or none.

When a Discovery run shows a background receipt, it reports what the run reviewed, what it created, and what needs attention. Use its links to open the right destination:

• Signals for source-backed evidence and its original context.
• Evidence for the atomic findings extracted from Signals and connected to related work.
• Insights for patterns synthesized from related Evidence.
• Ideas for imported requests, bets, or roadmap candidates.

Open the signal when you need to verify the text that arrived. For Pendo, use the integration result to distinguish NPS records from comment signals. If processing fails or the result looks thin, check the source body and status before retrying.

Curate the result by editing, merging, splitting, or linking it to existing evidence. Then move from evidence to action with Insights → opportunities and From customer evidence to initiatives.

For rollups and stable labels, use Taxonomy & themes. Keep PRDs, decision logs, and other living product documents in Workspace context.