Skip to main content

Overview

Health scores provide a composite view of customer health, but the category breakdown is where actionable insight lives. This page covers how to read scores effectively.

Reading the Score

The overall score (0-100) maps to a risk level based on your configured score buckets. By default:

Using the Category Breakdown

The overall score is a starting point. Click into the Health Score tab on any customer to see the category breakdown and identify which area is driving the score up or down. Read the analysis, supporting sources, assessment date, and evidence limitations alongside the score. Quivly can combine permitted evidence across sources, but an inferred value should remain distinguishable from an observed one. Call evidence and Slack evidence have different units and source context; a conversation count is not necessarily a message count. What to look for:
  • Lowest category score - This is where to focus intervention
  • Score vs weight mismatch - A low score in a heavily weighted category has outsized impact
  • Category with the most change - Use the trend indicator to spot which areas are shifting

On the Health Score tab, the history chart shows score changes over time:
  • Gradual decline over weeks suggests a real trend that needs attention
  • Sudden drop may indicate a specific event (e.g., spike in support tickets, missed renewal)
  • Score change at a version marker indicates a configuration change, not a change in customer behavior

When Scores Don’t Match Expectations

If a score doesn’t align with your knowledge of the customer:
  • Check for missing data - A disconnected integration, unavailable source, or stale usage can limit the assessment. Scoring can be skipped when evidence is insufficient, so check the last successful assessment date. Missing data is not evidence of zero activity or healthy behavior.
  • Review individual metrics - Use the Test tab or the score breakdown to see which metrics are contributing unexpected values
  • Consider timing - A new customer may have limited history. Check what the assessment actually observed before treating a low score as a confirmed problem.
  • Check the saved configuration - Read the bucket labels, criteria, source exclusions, and usage windows that belonged to that assessment. Current configuration edits do not rewrite historical scores.