Every store chat collects something most merchants never read: the questions it could not answer well. A shopper asks for a "waterproof jacket with a hood under 150 euros", gets nothing useful, and leaves. That single miss tells you more about your catalog than a week of page views. This post explains what unknown questions are, where they usually come from on a Magento 2 store, and how to turn them into fixes.
What counts as an unknown question
In a catalog-aware assistant, an unknown question is a turn where the shopper asked something the assistant couldn't ground in your data. Either no product matched, or the question was about something that isn't in the catalog or the store's curated knowledge at all. Because Claspwell answers from your synced Magento catalog rather than letting the model make things up, these gaps show up as gaps instead of being papered over with a confident, invented reply.
Where the gaps usually come from
- Missing or inconsistent attributes. Shoppers filter by things like material, size, compatibility or use case. If half your products have that attribute empty, or the values are spelled three different ways, the assistant can't match them.
- Words your catalog doesn't use. Customers say "sofa", your category says "couch". They ask for "trail shoes" when the attribute is "terrain: off-road". These are vocabulary gaps, not product gaps.
- Questions that aren't about products. Delivery times, returns, B2B terms, technical specs or installation questions that live in a PDF or in your support team's heads, not in the catalog.
- Real demand you don't serve. Sometimes the honest answer is that you don't sell it. Repeated requests for the same missing product are a buying signal worth passing to whoever runs the assortment.
How to turn them into fixes
A simple monthly routine works for most stores:
- Group the unknown questions by theme instead of reading them one by one.
- For attribute gaps, fill and normalise the attribute in Magento. The next catalog sync makes it available to the assistant. Our attribute checklist covers which attributes to fix first.
- For non-product questions, add a short, reviewed answer to the store's curated knowledge so the assistant can answer it the same way every time.
- For real demand you don't serve, send the list to merchandising. It is free market research from people who were already on your site.
Then check the same themes the following month. If a theme stops showing up, the fix worked.
Why this beats guessing
Search logs tell you what people typed. Unknown questions tell you what people typed in their own words, with context, and where your store failed them. Because the assistant asks clarifying questions before giving up, what is left over is usually a real gap rather than a vague query.
Where Claspwell fits
Claspwell is a catalog-aware AI sales assistant for Magento 2 and Hyvä. Answer analytics and unknown questions are included from the Growth plan up, alongside hourly catalog sync. Plan details are on the pricing page. We install the Magento module with you, typically in 3 days to 1 week once scope and access are ready. Request an integration to see how it would work on your store.