metrics-compare

Metrics Compare

Compare support performance between two time periods to identify improvements, regressions, and trends. Surface actionable insights from the data.

Process

Step 1: Parse Periods

Interpret the user's request into two concrete date ranges. Common patterns:

• "this week vs last week" — Current Mon-Sun vs previous Mon-Sun
• "March vs February" — Full calendar months
• "last 7 days vs previous 7 days" — Rolling windows
• "this quarter vs last quarter" — Q1/Q2/Q3/Q4 boundaries

Use the configured timezone (default: UTC) for all date calculations. If the user provides ambiguous ranges, ask for clarification before proceeding.

Step 2: Fetch Data

1. Fetch metrics for Period A using get_metrics with the first date range
2. Fetch metrics for Period B using get_metrics with the second date range
3. Fetch channel analytics for both periods using get_channel_analytics to break down by channel
4. Fetch agent performance for both periods using get_agent_performance to see individual trends
5. Optionally fetch trend data using get_trends for a continuous view across both periods

Step 3: Calculate Deltas

For each metric, compute:

• Absolute change: Period B value minus Period A value
• Percentage change: ((Period B - Period A) / Period A) * 100
• Direction indicator: Up arrow for increase, down arrow for decrease, right arrow for no change

Step 4: Assess Significance

Apply the significance threshold to highlight meaningful changes:

• Changes greater than +10% or less than -10% are flagged as significant
• Changes between -10% and +10% are considered stable
• For metrics where lower is better (response time, resolution time), a decrease is positive

Step 5: Present Comparison

Format the results as a side-by-side comparison table with insights.

Delta Calculation Rules

• Percentage change: Always calculate relative to Period A (the earlier period)
• Division by zero: If Period A value is 0, show "New" instead of a percentage
• Rounding: Round percentages to one decimal place
• Direction context: For some metrics, "up" is good (resolution rate, CSAT) and for others "up" is bad (response time, open conversations). Apply appropriate color coding in the assessment.