Notice a pattern
An aggregate count can show that one approved theme keeps ending in a knowledge-gap outcome.
Aggregate insights
See aggregate themes. Investigate the reason.
When reporting is enabled for your assistant, Roleline counts fixed topics and outcomes. The counts can point your careers team towards an area worth reviewing; they do not reveal the exact question or explain the cause.
What the report can show
Counts of approved themes, never conversation text.
Each exchange is classified against a fixed taxonomy your team approves and summarised as aggregate counts. Depending on what is set up for your assistant, that can include recurring topic themes, knowledge-gap outcomes, official-source routing and conversation-depth bands.
Messages may be processed to classify an exchange. The central reporting store keeps approved labels and counts, and client reports contain aggregate results rather than conversation text. How long conversations are kept, and what is logged where, are agreed and checked for your own assistant rather than assumed.
Fixed topicEligibility
Fixed outcomeKnowledge gap
Aggregate reportCount within the selected window
Illustrative structure only. No usage number is shown or implied.
How to use it
A signal starts an investigation. It does not finish one.
Review the source
Your careers team still needs to read the relevant content, or run an agreed evaluation set, to understand why.
Change deliberately
Any change to content or configuration goes back through review and evaluation before your team signs off a new version for release.
Questions for review
Ask what the data means before acting on it.
- Is reporting enabled for the assistant these counts came from?
- Which fixed taxonomy and outcome definitions produced the counts?
- Does an agreed evaluation set reproduce the apparent pattern?
- Who owns the source content the pattern points at?
- What evidence will show that a change helped?
Sample candidate insight report
Read a full sample report.
The sample below uses synthetic demonstration data for Example Retail Ltd (fictional employer), our demonstration employer, and is generated by the same code that produces a real client report each month — so it matches the aggregate report the current product can produce, including the rows it withholds when a count is too small to share safely.
Every number in it is invented. It contains no client data, exact candidate questions, automatic diagnosis or performance claim.
Open the sample reportA useful next step
Bring one real careers-site problem.
Tell us what candidates struggle to understand or find. In a free 20-minute fit call, we'll work through the problem with you and see how Roleline could help.