Metric Drivers: The starting point for understanding CSAT and deflection
Metric Drivers helps CX teams quickly understand what’s driving CSAT drops and escalation without digging through thousands of conversations.

Customer experience leaders have more data than ever before. Thousands of conversations pour in every day across chat, email, and voice. Dashboards track deflection rates, CSAT, and escalation volumes. But when the trends shift, where do you even begin figuring out what went wrong?
Most teams resort to manual QA, combing through endless transcripts to spot patterns. It’s slow, reactive work and by the time you’ve pieced things together, the customer pain has already spread.
With Metric Drivers, part of Decagon’s Insights & Reporting suite, finding the “why” behind your key business metrics is no longer a guessing game.
From data overload to direction
Metric Drivers uses AI to read every conversation that resulted in a poor CSAT or an escalation. Then, it identifies the underlying reasons customers got stuck and groups them into clusters of recurring issues most correlated with your metrics.
Unlike traditional analytics, which treat each conversation as a single datapoint, Metric Drivers can extract multiple reasons from the same conversation. That matters because support interactions are rarely black and white. A customer might be frustrated because their card didn’t work and because they weren’t offered a clear way to escalate. Metric Drivers captures both, giving you a more realistic view of what actually drives dissatisfaction.
At a glance, you’ll see your top drivers ranked by impact. You can:
- See the biggest drivers of CSAT drops or deflection without manual review
- Filter by tag or date to focus on specific cohorts or timeframes
- Click into any cluster to investigate granular drivers and related conversations
- Export results for deeper analysis or reporting

Think of it as a force multiplier for your team, giving them a clear starting point for investigation and iteration.
The AI-powered workflow
Surfacing issues is only the beginning. Metric Drivers is built to power continuous insights, action, and monitoring that help your AI agent improve over time:
- Insights: Metric Drivers analyzes escalations and low-CSAT conversations, spotting multiple reasons for dissatisfaction and clustering them into clear themes.
- Action: Each cluster helps you identify what to do next. For example:
- Found a cluster where refund waivers aren’t working? Update the Agent Operating Procedure (AOP) to include clear logic for when and how to apply them.
- Noticed a spike in login failures driving CSAT down? Flag it to engineering with concrete conversation evidence.
- Seeing repeated policy confusion? Update your knowledge base so both AI and humans give consistent answers.
- Monitoring: Metric Drivers inform your Watchtower flags to track if the issue resurfaces. You’ll know whether the changes stick and catch issues early if they come back.

Metric Drivers gives you the signal. AOPs and Watchtower help you take action and stay ahead of issues before they spread.
When AI needs your team’s help
Traditional chatbots built on rigid decision trees could only react, often trapping customers in frustrating loops. Large language models changed the game and AI agents can now resolve complex questions at scale. But even the best AI can’t solve everything.
Metric Drivers highlights where your AI agent fell short and where your team needs to step in to guide and improve it. As more support volume is handled by AI, CX teams are shifting from resolving individual tickets to managing and optimizing the systems behind them. Metric Drivers gives you the visibility and control to make that shift so your team can focus on high-leverage fixes, not high-volume reviews.
Customers have told us that the clusters surfaced by Metric Drivers confirmed what they suspected about their AI agent’s blind spots and even uncovered new opportunities to improve. That evidence gave them a stronger case for product fixes and sharper feedback loops with their teams.
See what’s behind the metrics
Metric Drivers helps you move beyond scores and surface metrics to understand what’s actually going wrong and where your AI agent needs an update.
Use it to uncover hidden gaps, sharpen your AOPs, and build smarter Watchtowers that keep issues from slipping through the cracks.
If you’re a Decagon customer, connect with your Decagon team to enable this within your Insights and Reporting dashboard. New here? Request a demo to see how our AI agent platform helps you turn insights into impact.
Metric Drivers: The starting point for understanding CSAT and deflection
September 3, 2025

Customer experience leaders have more data than ever before. Thousands of conversations pour in every day across chat, email, and voice. Dashboards track deflection rates, CSAT, and escalation volumes. But when the trends shift, where do you even begin figuring out what went wrong?
Most teams resort to manual QA, combing through endless transcripts to spot patterns. It’s slow, reactive work and by the time you’ve pieced things together, the customer pain has already spread.
With Metric Drivers, part of Decagon’s Insights & Reporting suite, finding the “why” behind your key business metrics is no longer a guessing game.
From data overload to direction
Metric Drivers uses AI to read every conversation that resulted in a poor CSAT or an escalation. Then, it identifies the underlying reasons customers got stuck and groups them into clusters of recurring issues most correlated with your metrics.
Unlike traditional analytics, which treat each conversation as a single datapoint, Metric Drivers can extract multiple reasons from the same conversation. That matters because support interactions are rarely black and white. A customer might be frustrated because their card didn’t work and because they weren’t offered a clear way to escalate. Metric Drivers captures both, giving you a more realistic view of what actually drives dissatisfaction.
At a glance, you’ll see your top drivers ranked by impact. You can:
- See the biggest drivers of CSAT drops or deflection without manual review
- Filter by tag or date to focus on specific cohorts or timeframes
- Click into any cluster to investigate granular drivers and related conversations
- Export results for deeper analysis or reporting

Think of it as a force multiplier for your team, giving them a clear starting point for investigation and iteration.
The AI-powered workflow
Surfacing issues is only the beginning. Metric Drivers is built to power continuous insights, action, and monitoring that help your AI agent improve over time:
- Insights: Metric Drivers analyzes escalations and low-CSAT conversations, spotting multiple reasons for dissatisfaction and clustering them into clear themes.
- Action: Each cluster helps you identify what to do next. For example:
- Found a cluster where refund waivers aren’t working? Update the Agent Operating Procedure (AOP) to include clear logic for when and how to apply them.
- Noticed a spike in login failures driving CSAT down? Flag it to engineering with concrete conversation evidence.
- Seeing repeated policy confusion? Update your knowledge base so both AI and humans give consistent answers.
- Monitoring: Metric Drivers inform your Watchtower flags to track if the issue resurfaces. You’ll know whether the changes stick and catch issues early if they come back.

Metric Drivers gives you the signal. AOPs and Watchtower help you take action and stay ahead of issues before they spread.
When AI needs your team’s help
Traditional chatbots built on rigid decision trees could only react, often trapping customers in frustrating loops. Large language models changed the game and AI agents can now resolve complex questions at scale. But even the best AI can’t solve everything.
Metric Drivers highlights where your AI agent fell short and where your team needs to step in to guide and improve it. As more support volume is handled by AI, CX teams are shifting from resolving individual tickets to managing and optimizing the systems behind them. Metric Drivers gives you the visibility and control to make that shift so your team can focus on high-leverage fixes, not high-volume reviews.
Customers have told us that the clusters surfaced by Metric Drivers confirmed what they suspected about their AI agent’s blind spots and even uncovered new opportunities to improve. That evidence gave them a stronger case for product fixes and sharper feedback loops with their teams.
See what’s behind the metrics
Metric Drivers helps you move beyond scores and surface metrics to understand what’s actually going wrong and where your AI agent needs an update.
Use it to uncover hidden gaps, sharpen your AOPs, and build smarter Watchtowers that keep issues from slipping through the cracks.
If you’re a Decagon customer, connect with your Decagon team to enable this within your Insights and Reporting dashboard. New here? Request a demo to see how our AI agent platform helps you turn insights into impact.