Conversation insights
Aggregate dashboard numbers only tell you so much. For what's actually happening inside individual conversations, ShopGuide gives you two tools: the transcript viewer, and an automatic detector that flags where the AI didn't fully help.
The transcript viewer
Open Chats and you'll find every conversation for your store, each one readable in full. Each transcript shows:
- The complete message history between customer and AI
- Tool badges — which tools the AI used (for example Search, Cart, Checkout, Discount, Order, Support, Reviews, Loyalty), so you can see what actions were taken, not just what was said
- Customer feedback — thumbs up or down on individual replies. A thumbs-down asks for a short explanation first (at least 50 characters), so negative ratings come with context instead of just a number
- Filters by rating, tool used, date range, and whether the session had any interaction at all
This is the most direct way to read real conversations instead of relying on aggregates.
Knowledge Gaps detection
After every completed response, ShopGuide runs a quick background check — never blocking the chat — to see whether the AI failed to fully resolve what the customer needed. If it did, the gap is classified and stacked with other occurrences of the same underlying question, so "how is this packaged?" and "what's the shipping box made of?" count as one recurring gap instead of two separate entries.
The Missing Information card ("Information that the agent couldn't find when customers asked.") is organized by what you should do about it:
| Tab | What it holds |
|---|---|
| All | Every open item, with a count |
| Add product info | Your product pages don't answer the question — packaging, materials, sizing |
| Clarify a policy | A policy question your store content doesn't answer |
| Customer demand | Customer requests ShopGuide couldn't act on |
| Fix your store | Something customers reported as broken in the store |
Rows have a plain-language headline, a severity badge (high, medium or low) and a context badge (Product, Page, Policy or General). They are ranked by lost sales, not just frequency. Each row shows how many of the conversations did not end in an order; once a row has at least 10 conversations it also shows the share that bought against your store average, and rows well below your average get a Below your average badge. This ranking only appears when your store has chat-attributed orders in the period, and it never claims a money amount.
Row actions: View Chat (or View N Chats), resolve or dismiss, and on "Add product info" and "Clarify a policy" rows Answer in Knowledge Base. That opens a form pre-filled with the customer's phrasing, saves a custom FAQ into Shopify's own Knowledge Base app and marks the gap resolved. ShopGuide's chat reads your published Knowledge Base FAQs automatically (changes can take about 10 minutes to appear), and you can edit them on Training in the Shopify Knowledge Base card. It's a detection and triage tool, not an automatic fix: deciding what the answer is stays with you.
Sentiment analysis, emotion detection, topic clustering, churn prediction — none of that runs here. What's actually generated from your chat data is tool usage, thumbs up/down with written feedback, and categorized, topic-stacked knowledge gaps. That's it, and it's usually enough.