Traits screen
Tone tells you whether assistants speak well of a brand. Traits tell you what they say that brand is for.
Individual statements are grouped into these traits - "opaque pricing", "strong accessibility expertise" - and Traits explains where they come from.
The report has two tabs: Overview puts every brand on one grid, Profile reads one brand at a time.
Overview
On the Overview tab, traits run down the side and brands run across the top, with criticisms in their own band at the bottom. The grid opens sorted by your own column - your strongest trait first - and any header re-sorts it.
Each cell is that brand's share of its own responses that discussed the trait - not its share of the row. That is what lets you compare two cells on one row directly even when one brand is named ten times as often as the other.
- An asterisk marks the clear leader on a row: highest, and beyond the margin of error.
- Discussed, never said means the brand was analyzed and the trait never attached to it. A real result.
- A hatched cell means the trait never came up, or there were too few responses to say anything. An absence of data, not a zero.
Clicking any cell opens the claims behind it: the actual statements, what kind each was, which prompt produced it, and which page the assistant cited alongside it.
Profile
The profile is one brand's ranked traits - what assistants reach for when that brand comes up, most often first - split into Known for and Criticized for, so praise and complaints are never mixed on one scale. Above the bars, the report calls out its strongest findings: a criticism the brand owns outright, a strength the workspace grants that assistants rarely mention, or a clear lead over every rival.
It defaults to your own brand, and the Profile picker in the filter row switches it to any brand you track. Looking at a rival's profile is reading a page about them; it changes nothing anywhere else.
A line above the bars says what the profile was read from:
- A brand with brand topics measured in the window is read from the answers to questions about it directly ("tell me about X"), with the workspace's answers alongside for comparison.
- Any other brand - most rivals - is read from what the workspace's answers say about it.
If a brand has no brand topics, adding some is what unlocks the direct reading.
What you can do here
- Drill into any trait to read the supporting claims.
- Hide a trait that is noise from the row menu. It keeps matching claims quietly so the same trait is never re-proposed, but stays out of reports.
- Manage traits - merge duplicates, restore something you hid, flag a trait as a criticism (or unflag one, for the workspace where "expensive" is a boast), or ask Silktide to look for new traits in the ungrouped claims.
The trait list is deliberately capped. When an arena is full, merge or hide something before finding more; a short list is what keeps this readable.
If it is empty
Traits need enough claims to see patterns, and grouping happens automatically as batches finish, so a new workspace will have nothing here for a while. Widening the date range is the other common fix.
Related
- Traits - the concept, and why criticism has an address
- Reputation screen - how favorably those traits are said
- Sentiment - the four states behind these judgments
- Citations screen - the sources behind the claims