Monitor performance in Klaviyo
Once Customer Agent is live, performance data helps you understand more than whether the agent is answering questions. It shows how conversations are resolving, how customers are engaging after those conversations, and what kind of business impact the agent is driving, so you can spot what’s working, identify what needs attention, and know where to optimize next.
Click the gallery tabs below to explore the main areas in Klaviyo where you can find actionable Customer Agent data.
Customer Agent analytics page
The Customer Agent analytics page (Customer Agent > Performance > Analytics) houses all of your important and relevant Customer Agent data. Here’s what you can see:
- A chart providing total chat volume
- You can adjust filters by time period, and by channel.
- Key metrics cards, showing specific data for:
- Total resolved by AI
- Percent resolved by AI
- AI generated sales
- Add to carts
- Link clicks
- Average order value
- Skill performance (how many times each skill is used, filterable by channel)
- Conversations (including data for the question asked, channel, status, tags, time sent)
Metrics tab
Certain Customer Agent event data can be found in your general Metrics tab (Analytics > Metrics). You can use this data to create new segments or trigger flows. Events you can find in the Metrics tab are:
- Started Conversation (when a shopper starts a new Customer Agent conversation)
- Ex. Use this event to trigger a follow-up flow for shoppers who started a conversation with Customer Agent, but may need an extra reminder or nudge to complete their purchase.
- Conversation Resolved (when Customer Agent resolves a conversation, no escalation needed)
- Ex. Use this event to trigger a post-conversation flow, such as a satisfaction check-in or a follow-up message with helpful next steps after Customer Agent resolved the issue.
- Conversation Escalated (when Customer Agent needs to hand off the conversation via escalation route)
- Ex. Use this event to trigger a temporary pause or exclusion from promotional flows, so shoppers with an active support issue don’t keep receiving marketing messages while they’re waiting on a human follow-up.
- Received Product Recommendation (when Customer Agent surfaces product recommendations in the chat)
- Ex. Use this event to trigger a personalized follow-up featuring the recommended products, helping you continue the shopping journey after the chat ends.
Profile Activity Feed
If you’re looking to investigate or review data about a particular profile, you can see certain Customer Agent events in their Profile Activity Feed (Audience > Profiles, click into a given user’s profile, view profile activity feed).
You can see the same events per the Metrics tab, for that specific user:
- Started Conversation (when they started a conversation with Customer Agent)
- Conversation Resolved (when Customer Agent resolved their conversation, no escalation needed)
- Conversation Escalated (when Customer Agent needs to handoff their conversation via escalation route)
- Received Product Recommendation (when Customer Agent surfaces product recommendations to this user in the chat)
Interpret key Customer Agent metrics
Let’s look a little bit deeper into the key metrics that are surfaced in the Customer Agent analytics page. This data gives you valuable insight into what’s working, what’s not, and where you may focus next. In the table below, review what each metric tells you, and strategies for improvement.
When you make adjustments and improvements, preview and test your Agent before setting your changes live. Run simulations using the preview tab to ensure the Agent is answering as expected.
Metric | What it tells you | If it's low, improve it by... |
|---|---|---|
Total resolved by AI | How many conversations the agent fully handled without escalation and human intervention. | Add missing knowledge, improve weak content, and create or refine skills for common repeatable requests. |
% resolved by AI | How efficiently the agent is resolving conversations overall. For your % resolved by AI rate, compare your brand to benchmarks to see where you fall: Below average: Below 50% Good: Between 50-70% Best-in-class: 70%+ | Review escalations and failure patterns, tighten guidance, and improve handoff logic only where needed. |
AI-generated sales | Whether the agent is contributing to revenue, not just support. Web chat attribution uses a 24-hour window. Text attribution follows your Klaviyo text message attribution settings (KAV). | Improve product recommendation flows, make product guidance more specific, and ensure the agent links shoppers to the right products at the right time. |
Add to carts | Whether conversations are moving shoppers closer to purchase. | Strengthen recommendation quality, add clearer shopping guidance, improve the agent’s ability to guide shoppers toward a next step. |
Link clicks | Whether customers are engaging with the links and suggestions the agent sends. | Make responses more actionable, tighten CTAs, and ensure links are highly relevant to the question being asked. |
Average order value | Whether agent-driven orders are meaningful in value, not just count. | Improve upsell/cross-sell guidance, bundle recommendations, and higher-intent recommendation moments. |