Start with the facts
Sundial uses governed definitions, so everyone starts from the same facts.
An Analysis Agent for questions
that deserve real analysis.
Ask anything from a quick metric lookup to a complex investigation. Sundial grounds every answer in your company context, applies expert analytical methods, and shows you why the result can be trusted.
A lookup tells you what happened. Analysis explains why.
Sundial handles both, from a trusted number in seconds to a multi-step investigation that follows the evidence.
Investigate what drove them
Sundial plans the analysis, tests likely drivers, checks the data, and explains the evidence.
A complete analysis,
from question to trusted answer.
Sundial chooses the right method, grounds the work in your business, tests the likely drivers, and shows why the result can be trusted.
Understand the question
Sundial understands what you are asking and chooses the right analytical approach.
Sundial understands the request and selects the Metric Change playbook.
Trusted answers,
ready for the whole team.
Turn an analysis into a dashboard, report, or recurring update your team can keep using.
Measure answer quality with Observability and Evals.
See how every answer performs, identify where context or playbooks fall short, and validate improvements before they reach the wider team.
Find quality gaps with Agent Observability
Track question types, trust levels, playbook usage, and individual conversations. Drill into weak answers to see exactly what needs attention.
Catch regressions with Evals
Test governed lookups and deep analyses after changes to context, models, or the semantic layer. Compare quality before rolling changes out.
I used to write SQL and build charts across multiple tools to understand user behaviour. With Sundial’s analysis agent, what took hours now takes a single prompt. I get to decisions faster than ever.
Long Yin Luk
Product Manager, HEYMAX
Questions about analysis with Sundial
Is Sundial's Analysis Agent just text-to-SQL?
No. Sundial starts with governed semantic measures and dimensions whenever they fit the question. For deeper investigations, it can follow expert playbooks, combine multiple analysis steps, and use SQL or Python when the work requires them.
What questions can the Analysis Agent answer?
It handles governed lookups and trends as well as deeper questions about metric changes, experiments, funnels, retention, anomalies, user behaviour, and the drivers behind business performance.
How do playbooks work?
Playbooks encode repeatable analytical methods. When a question matches one, Sundial follows its steps, such as checking expected movement, decomposing a metric, testing explanatory segments, and ruling out data issues, rather than improvising a one-off approach.
How does Sundial understand our business?
The Analysis Agent draws from your semantic definitions, organization context, warehouse metadata, lineage, and connected knowledge sources. Data teams manage this context centrally so answers stay consistent with how the company operates.
How is confidence determined?
Every answer receives a high, medium, low, or unverified trust signal based on how it was produced. Sundial considers whether it used governed definitions and, when relevant, whether it selected and followed the appropriate playbook.
Can our data team inspect the work?
Yes. Answers can show the analysis steps, supporting definitions, playbook usage, sources, and custom queries. Organization-wide observability helps data teams review weak answers and improve the context behind them.
What can we do with the result?
Results can remain in the conversation as text, charts, and tables, or become shareable Artifacts and refreshable Data Apps. Depending on your workspace, teams can also export, version, schedule, and deliver analysis through Slack.
Put a world-class analyst behind every decision.
Give operators faster answers and data teams the context, methods, and control to trust them.