The Drill Insights agent helps you interpret the results of any Drill query and guides your next steps in analysis. It activates on the Drill page whenever you run a data drill-down query (either one you built manually or via Cee). After the query execution, a “Drill Insights” button becomes available on the Drill results screen. Clicking this button triggers Cee to analyze the returned data, extract key findings, and automatically post a summary of insights in the Cee chat panel. These insights go beyond the obvious graph outputs, highlighting what the numbers might indicate about user behavior or product performance, and suggesting further analyses to dig deeper:
- Key trends or anomalies observed (e.g. a sudden spike or drop on a certain date, an upward trend over the last weeks, etc.).
- Notable segment patterns (e.g. one user segment far outperforms others in the data, or a combination of attributes stands out as significant).
- Possible explanations or implications of these patterns, phrased as business insights (for example, linking a spike to a feature launch or identifying an underperforming user cohort).
Along with the narrative analysis, the Drill Insights agent provides a set of actionable insight buttons - clickable suggestions for what to do next. These suggestions are context-aware follow-up queries or actions, limited to core Countly features (Drill, Funnels, Cohorts) that Cee can assist with. For instance, after analyzing a Drill result, the agent might suggest “Create Cohort of Users in Segment X”, “Build Funnel for Event Y to Conversion”, or “Drill Down Further by Property Z”. By clicking one of these Cee will execute the next analysis or set up the respective query for you.
Analytical Techniques Employed
The Drill Insights agent leverages several data analysis techniques to interpret your query results and surface meaningful insights:
- Correlation Analysis: The agent looks for relationships or associations in the data. If your Drill query involves multiple variables or segments, it will measure how they move together or differ. In simple terms, correlation analysis “measures the strength and direction of relationships between two or more variables,” helping uncover patterns that might link user attributes to outcomes. For example, if you segmented an event by user demographics, the agent might detect that one segment (e.g. users on a certain app version or from a particular country) consistently has higher values, suggesting a meaningful correlation between that attribute and the event’s frequency. In multi-breakdown queries, it can even spot specific combinations of properties that are noteworthy. For instance, a particular device-model and app version pair with exceptionally high engagement compared to others, indicating a unique correlation.
- Trend & Anomaly Detection: When your Drill query has a time-series output (e.g. daily or weekly counts), the agent performs trend analysis to determine the overall direction and velocity of change, and flags any anomalies (outliers). It identifies if metrics are rising, falling, or cyclical over the selected period, and notes points that deviate sharply from the baseline. The agent uses this to highlight unusual spikes, dips, or plateaus in your data that might signify something noteworthy (e.g. a sudden drop in user sessions on a specific day could indicate an incident or change worth investigating). Trend analysis might reveal, for example, that an event is gradually increasing week-over-week, or that there’s a seasonal usage pattern (weekday vs weekend behavior), whereas anomaly detection will point out days where the metric was abnormally high or low compared to typical values. These findings help you quickly spot issues or opportunities without manually crunching the numbers.
- Seasonality & Changepoint Analysis: For longer time spans or repetitive patterns, the Drill Insights agent examines seasonality and detects any changepoints in the data’s behavior. The agent will check if your query data shows periodic fluctuations. For example, higher activity every weekend, or dips every holiday and will inform you of such regular patterns. Meanwhile, changepoint analysis is used to find moments in time where the underlying data distribution shifts significantly. In other words, it hunts for structural breaks in the time-series, points where the average or trend of the metric changes abruptly. The Drill Insights agent will flag such points (e.g. “On July 15, the daily purchase count jumped to a new higher range, indicating a potential changepoint – perhaps due to a feature release or campaign”). By catching changepoints, the agent alerts you to events after which user behavior is notably different. This can be crucial for correlating metric changes with product updates or external factors.