Conversations
What is the Conversation Explorer?
- Purpose: A tool for exploring and accessing complete conversation data, available by clicking each conversation.
- Value: The ability to access full conversation transcripts enables qualitative analysis of customer interactions, understanding of conversation context, and identification of specific failure points or success moments within individual exchanges.
- Scope: Part of Analytics
Key Concepts
The Conversation Explorer captures the complete customer journey from conversation initiation (whether by the business or the end user) through automation attempts, potential escalations, and final outcomes, providing end-to-end visibility into the customer service process.
The table displays the conversation start time, user identifier, channel, and provider. It can be filtered using multiple criteria, such as nodes visited during the conversation, detected Smart Intents, or language, among others.
If needed, the conversations can also be exported, with the CSV file showing more statistics.
Examples or Use Cases
A customer service manager at an e-commerce company notices increasing customer complaints about their chatbot being unhelpful during the holiday season.
Using the Conversation Explorer, they can apply multiple filters to investigate: they filter by date range, escalation status, provider, and meaningful interactions to identify conversations that failed to provide value and required human intervention. The filtered results reveal 847 conversations matching these criteria, showing a 40% increase compared to the previous month.
By examining the aggregate data, they discover that most failed conversations occurred during peak hours (2-6 PM) and involved customers asking about delivery delays and return policies. Checking the exchange of messages between the enduser and the bot, they noticed that the bot's responses about delivery tracking were outdated and didn't account for holiday shipping delays, while return policy explanations were too generic and didn't address gift return scenarios.
Best Practices
Use filters to subset the whole data.