Conversational analytics
How the data team learns what the business asks, and improves
A data team can only add the definitions it knows are missing. Mark Your Data Assistant logs what the business asks, whether each answer was governed and why users gave a thumbs down, so the data team can close each gap in dbt.
How it works out of the box
Every question is a signal. The log shows what colleagues ask, and which answers were governed or unverified.
A thumbs down has a reason. In Teams, a thumbs down requires a reason, and each reason points to a different place to look.
Gaps close in dbt. The data team adds a metric or saved query through a normal pull request, with no prompt tuning.
The coverage scan shows the change. The scan asks a list of questions and reports each one as governed, unverified, asked back or refused.
Correctness is checked separately. ✅ Governed means answered from your definition. The data team checks correctness with its own reference questions.
What it means for you
- Analytics engineer: You get a backlog of definitions based on real questions.
- Manager: You can follow the share of governed answers over time.
- Business user: Your feedback reaches the data team, along with the answer it's about.
Mark Your Data calls this way of working Question-Driven Development.
Questions and answers
How do we find out what the business actually asks?
The log in Mark Your Data Assistant shows what the business asks, whether each answer was governed, and the reasons behind each thumbs down. The data team reads it through log reports and its own observability tool, and can export the questions for review, with user IDs left out unless requested.
What does a thumbs down tell the data team?
A thumbs down in Teams tells the data team where to look, because it requires a reason. "Data is missing or incomplete" points to a coverage gap, "The numbers look wrong" to the metric or SQL, and "My question was misunderstood" to routing or naming. Mark Your Data Assistant stores the reason with the exact answer it rates.
How do we close a gap the assistant found?
Add a metric or saved query in dbt through a normal pull request that is reviewed and tested. Once that change is published, Mark Your Data Assistant answers the same question from the new definition and labels it ✅ Governed, for everyone. Your data team closes the gap without any prompt tuning.
What is Question-Driven Development?
Question-Driven Development is a way of working for dbt teams in which the questions colleagues actually ask act as the test suite for the semantic layer. An unverified answer is a failing test, and a metric or saved query added in dbt makes it pass. Mark Your Data Assistant labels the answers and measures the coverage.
How do we show management that the answers are getting better?
Show the share of governed answers over time. The coverage scan in Mark Your Data Assistant asks a fixed list of questions before and after a dbt change and reports the difference per question. Governed means answered from your definition, so the data team checks correctness separately with reference questions.