Conversation Insights
Shoppers tell the AI what they wanted and why they hesitated. Conversation Insights reads the last seven days of shopper conversations across web chat, email and messaging channels and groups what keeps coming up, so you can act on it instead of reading transcripts one by one. It is included from the Growth plan up and needs the analytics permission.
Where it lives
Open Analytics → Conversation Insights in the dashboard sidebar, or go to /app/<storeId>/insights. A new analysis runs once a week for the previous seven full days (UTC). You can also start a refresh from the page, at most once every six hours per store.
The four theme lists
- Product gaps - products, sizes or features shoppers asked for that your synced catalog does not appear to carry.
- Objections - reasons shoppers hesitated, such as price, shipping, returns, sizing, or trust.
- Competitor mentions - other stores, brands or marketplaces shoppers compared you with.
- Top questions - what shoppers ask most, each a candidate FAQ.
Each theme shows how many distinct conversations raised it in the period, the trend against the previous period, a few example quotes with personal details removed, and links to sample conversations. Counts are calculated from the conversations themselves, not estimated by the AI.
Acting on a theme
- Open a sample conversation and confirm the shopper actually raised the theme.
- For a product gap, use Check catalog: the product may exist under a name shoppers do not use, which is a catalog copy fix, or it may be a real gap worth stocking.
- For an objection, use Add objection response to open objection rules under Rules & Voice.
- For a question or comparison, use the Draft FAQ action on the theme. It creates a draft in the AI Improver review flow; nothing reaches shoppers until you complete and approve it.
When the page is empty
Related
- Analytics & weekly review - the numbers side of the same week.
- Training - knowledge gaps and the AI Improver review queue.
- Conversations - search every transcript.
Example: turn a theme into a checked answer
An illustrative theme says shoppers ask whether a jacket fits over a sweater. Open the linked conversations and confirm that pattern. Check the size guide and product measurements, then use Draft FAQ only if those sources support an answer. Review the draft, wait for indexing after approval, and test the original question. The theme count comes from conversations; the AI’s interpretation still needs your review.