Office Report Drilldown

Office Report Drilldown

Use this skill when the user has a headline number and needs the path from summary to drivers.

Use when

• The task starts from a KPI or management report.
• The user wants to know what changed, where it changed, and what likely drove it.
• A layered drill-down is more useful than a flat descriptive summary.

Do not use when

• No top-line metric is defined.
• The source data is too weak to support segmented analysis.
• The user only wants formatting help on the report.

Drill-down principles

• Start from the reported KPI, not from random slicing.
• Move from top line to segment to driver in a controlled sequence.
• Distinguish variance explanation from root-cause proof.
• Do not present correlation as causation.

Workflow

1. Lock the metric definition

Clarify:

• exact KPI
• period and comparison basis
• numerator and denominator if applicable
• whether the report reflects booked, billed, recognized, forecast, or estimated numbers

2. Find the first useful cut

Choose the highest-signal split, such as:

• business unit
• region
• product line
• channel
• customer segment
• time bucket

Use the split that best explains movement, not the one that is easiest to compute.

3. Quantify variance contribution

Break the movement into:

• biggest positive contributors
• biggest negative contributors
• mix effect vs volume effect when possible
• anomalies that need separate explanation

4. Test the likely driver path

For each major movement, ask:

• is this broad-based or concentrated
• is it one-off or recurring
• does it align with known business events
• what evidence is still missing for a stronger claim

5. Close with decision-ready language

The final answer should say:

• what moved
• what most likely drove it
• what remains uncertain
• what follow-up analysis would resolve the uncertainty

Output format

Return:

1. KPI frame

• metric
• comparison period
• source assumptions

2. Drill-down path

• segment cuts used
• why those cuts matter

3. Main drivers

• ranked contributors
• anomalies
• confidence level

4. Recommended next query

• one or two targeted checks for stronger root-cause confidence

Use together with

• office-duckdb-query
• office-source-reconciliation
• office-briefing