Review the Q&A Log

The Q&A log at /admin/qa-log is every question your team has asked, paired with the answer Muninnbase returned and the documents it cited. It is the closest thing the product has to "what is my team actually asking?" Read it weekly during the first month and you'll learn more about your knowledge base than from any other surface in the admin panel.

Filters

Three filters at the top of the page. Date range (open-ended start and end; either side can be empty), status (Answered, Unanswered, or All), and search (matches against the question text). They combine freely. "Unanswered in the last seven days containing the word 'PTO'" is a normal query and a very common one.

The default view is unfiltered and sorted by most recent. Most weeks that's exactly what you want; start there and narrow as needed.

What each row tells you

Question (the actual text the user typed), status (Answered / Unanswered), response time (in milliseconds), a feedback icon if the user reacted with thumbs up or thumbs down, and timestamp. Click Review on any row to open the side detail panel.

The detail panel

The panel shows the full answer the user got back, the source citations (filename, section, page) that the answer was built from, the response time, and the similarity score. The similarity score is a number between 0 and 1; it indicates how confident retrieval was in the chunks it returned. Higher is more confident, with most real answers landing somewhere between 0.4 and 0.8.

If the user thumbed the answer down with a comment, the comment appears below the answer in a blockquote. The panel is read-only. You cannot edit a historical answer, and you shouldn't want to. The log is a record of what actually happened.

Exporting to CSV

The Export button at the top right downloads the currently-filtered view as a CSV file. Useful for sharing a week's worth of unanswered questions with whoever owns the relevant policy, archiving before the retention purge runs, or doing your own analysis in a spreadsheet.

The export respects every filter you have applied. If you filter to "unanswered in May with the word 'expense' in the question," that's the CSV you get.

Most common questions

Below the log table, a "Most common questions" panel ranks the top 10 questions by frequency, scoped to your tenant and respecting the same date filter as the log above. The aggregation is forgiving about capitalization and punctuation, so "How does PTO accrual work?" and "how does pto accrual work" land in the same bucket.

Treat this list as your editorial priorities. The questions employees ask most often are the questions your documentation should answer most clearly. If "PTO accrual" is at the top of the list every month, the answer to "how does PTO accrual work" had better be the cleanest paragraph in your handbook.

A note on anonymity

The log does not show who asked what. Questions and answers are logged; the user identity is not stored on the row. This is by design. Employees should be able to ask questions about their own policies without worrying that "asking" itself shows up in a manager's audit later. You'll see what your team is asking, but not who asked it.

This also means that the log survives the people in it: removing a user does not remove their past questions, because those questions were never tied to their name in the first place.

A note on retention

The log only keeps as much history as your tenant's retention window allows. Default is 30 days; the dropdown at /admin/settings offers 7 days, 30 days, 90 days, 1 year, or Keep all. Shortening the window deletes older entries the next time the retention job runs (the set up your workspace article covers the dropdown and the warning that appears before it applies).

If you want longer lookback for trend analysis, lengthen the window. The unanswered events queue is separate and is never auto-purged.

What to do next

The log shows you what your team asked; the unanswered events queue is where you act on the gaps it reveals. Resolve unanswered questions covers the workflow that turns the log from a reading exercise into a documentation improvement loop.