A mention is a single piece of customer feedback once it lands in Pilea, ready to be linked to a customer, grouped into themes, and turned into backlog work. One support ticket, one Slack message, one survey answer - each becomes a mention.

If backlog items are the conclusions, mentions are the evidence.

Where mentions come from

Mentions arrive from the sources you’ve connected: Slack, HubSpot, Attio CRM, Intercom, Zendesk, Notion, Support Ticket, Manual, Import, and Survey. Call recordings synced from Gong, Fireflies, Jiminny, Granola or Attio, and files you upload yourself, arrive as imports. Every mention keeps its source, so you always know where a piece of feedback originated.

What happens to each mention

Every mention goes through the same pipeline:

PII redaction (if enabled) - personal details are removed before anything else touches the text. Analysis waits until redaction has confirmed the text is safe.
Feedback cleaning (optional, per workspace) - long feedback such as tickets and transcripts is condensed to its product-relevant essence before analysis. Turn it on under Settings → Analysis, with optional instructions for what to keep. Short feedback is analyzed as-is.
Sentiment analysis - the mention is scored as Positive, Negative, Neutral, or Mixed.
Classification - the mention is sorted into your framework’s analysis units.
Customer linking - the mention is connected to the customer it came from.
Item linking - the mention is linked to matching backlog items, or becomes the seed for a new one. Each link records a verbatim excerpt, the sentence that justified it.

Very long mentions are analyzed in context-preserving chunks instead of being cut off.

You can watch this happen. Mention cards show their current state:

State	What it means
Waiting for redaction	Queued for PII removal
Waiting for analysis	Redacted, queued for analysis
Processing	Analysis is running now
Processed	Done - linked and classified
No items found	Analyzed, but nothing in it matched or produced a backlog item

ℹ️ Note: “No items found” isn’t an error. Some feedback is a thank-you note or small talk - not everything needs to become work.

Reading a mention

Long mentions are truncated in the list; click Read more to expand one in place. If feedback cleaning condensed the text, the list and the mention detail show the cleaned version with a Cleaned for analysis badge. Click Show original to see the customer’s exact words. The original is never modified.

Open a mention to see its linked Backlog items, Customers, and Tags. You can add or remove any of them by hand if the analysis got it wrong.

Finding mentions

Use search alongside filters for Source, Date, Customer, and Type (analysis unit). Sort by Newest first or Oldest first. Toggle Hide mentions without backlog items to drop the “No items found” mentions from the list.

💡 Tip: Filter by Customer before a big account’s check-in call. Reading their raw mentions in order is the fastest way to reconstruct what they’ve actually been telling you.

Actions on a mention

Archive - for test data, internal chatter, or the one customer who spams every channel. Archived mentions disappear from the default list, the overview numbers, and every backlog, tag and customer counter, while the source data stays intact. Click Archived in the toolbar to see them, and Restore to bring one back. Select several to archive or restore in bulk.
Move to another workspace - removes the mention from this workspace and re-analyzes it against the destination’s backlog. Tags and customers come along. Survey responses can’t be moved.
Delete - removes the mention.
Bulk delete - select multiple mentions and remove them in one go, handy after a messy import.

Re-analyzing

After you’ve refined your company context, analysis context, cleaning settings or framework, you can run the analysis again so existing mentions benefit from what Pilea now knows.

Whole workspace - go to Settings → Analysis and click Re-analyze workspace. Pilea removes the previous analysis output (backlog links, tags, customer links, sentiment and classifications, including ones added by hand), then re-runs every mention in the background. A progress banner shows how far it has got, and you can leave the page. You can pick a different Analysis framework as part of the run; switching framework also rebuilds the backlog items, since they carry the old framework’s types. Archived mentions and survey responses are skipped. Organization Admins and Owners only, once per 24 hours.
A selection - re-analysis of chosen mentions is available through the API and MCP.

From mention to backlog

You don’t manually promote mentions - Pilea links them to backlog items as part of analysis. The mentions attached to an item are what drive its evidence: mention count, customer count, sentiment, and recency all trace back to individual mentions. When you’re skeptical about an item’s score, read its mentions. The receipts are all there.

Next: Managing your backlog

Working with mentions