Conversational analytics for ecommerce lets you ask your store, ad and accounting data a question in plain English and get the figures back, instead of building a dashboard or exporting to a spreadsheet. In Luca you ask in the chat, typed or spoken, and each answer comes back with figures, a chart where useful and next steps, with every figure linked to the data table it came from. Luca answers about your own business data, not your customers' conversations.
Elsewhere, conversational analytics often means analysing contact-centre calls, support tickets and chatbot logs. Luca does none of that. It answers from the sources you have connected, as they stood at last night's refresh. For the wider case for asking instead of reporting, read the blog post on conversational analytics for ecommerce.
At a glance
| How you ask | Type in the chat box, which reads Ask Luca about your business, or speak through voice input. Follow-ups stay in the same thread, and Luca asks back when a question is ambiguous. |
|---|---|
| Sources one answer can use | Every source you have connected, such as Shopify, Amazon Seller Central, Meta, Google Ads and Xero |
| What an answer contains | Figures, a chart where useful and next steps |
| How you check it | Every figure links to its data table; checking a figure shows how to trace it |
| How fresh it is | Data refreshes once a day, overnight, not in real time |
| Question limits | None: unlimited questions and no AI credits |
| Where you can ask | The Luca chat, and Slack on the Leadership plan and above |
What you can ask, by kind of question
Each row is one kind of question, an example you could type and the page that covers it. Pages marked as planned are not written yet.
| Kind of question | Example question | Where it's covered |
|---|---|---|
| A question about one source | "What were our Shopify net sales by day last week?" | Ask a question |
| A question across several sources | "Did Amazon fees eat more of our margin than Shopify's this quarter?" | Answers across your sources |
| Why a number moved | "Why did our average order value fall in October?" | Ask why a number moved |
| Checking a figure | "Which table does this revenue figure come from?" | See the data behind an answer |
| A chart | "Chart weekly orders against Meta spend for this year." | Charts (planned) |
| A file the sources don't hold | "Compare this supplier price list with our product margins." | Attach a file (planned) |
| Market or benchmark context | "How does our repeat rate compare with typical baby-products brands?" | Web research (planned) |
| A what-if | "What happens to profit if ad spend rises 20%?" | What-if questions (planned) |
Luca shows question cards matched to the platforms you have connected, plus a library of about 50 expert questions you can pin. Those cards are the quickest way to see what a good question looks like for your own data.
General business questions
The same chat answers general business questions too, such as an HR or compliance question, or a request for content. You might ask how to structure a probation review, what a returns policy usually covers, or for a product description for a new sleep bag. Luca also drafts on-brand written content, and its image generation produces ads, posters and social images in your brand's colours, fonts, logo and voice. You refine either by replying in the thread.
Know where these answers change character. A question about your orders, ad spend or ledger is answered from your connected data, and each figure links to its table. A general question is answered as general guidance, not from your data, so it carries no source table to check. Treat HR and compliance answers as a starting point, not legal or HR advice, and take anything binding to a qualified adviser. For more on what Luca drafts alongside your numbers, see the guide to AI agents for ecommerce.
Where to start
Pick the role closest to yours and ask its first question. Each one points to the page that shows you how to take it further.
Founder
Ask "How did we do last week?" and read the figures, the chart and the next steps that come back. Then follow up in the same thread, as the Ask a question page in the table above explains.
Finance lead
Ask "Do last month's orders, payouts and the Xero ledger agree?" Luca joins the store and the books in one reply, and answers across your sources shows how to word the join.
Head of growth
Ask "Did Meta and Google spend leave us more profit this month than last?" The answer sets spend beside store revenue and costs, as the Answers across your sources page explains. For break-even targets, the page on ad profitability goes deeper.
Operations lead
Ask "Which best sellers run out first?" and follow up on the products it names. The question cards on the Ask a question page give you more stock questions, and reorder points cover when to order.
Common questions
Is my voice kept when I speak a question?
No. When you use voice input, Luca transcribes your question into the chat box, and the audio is not stored. Read the transcribed text before you send it, so a misheard product name doesn't change the answer.
Can my whole team ask questions in the same organisation?
Yes. Plans include 1, 3 or unlimited seats, and each colleague you invite takes a seat on your plan. Everyone asks from the same connected data, so answers agree.
What is the difference between chat and the Numbers dashboard?
Chat answers the question you ask, when you ask it. The Numbers dashboards are built for each brand by the Luca team and refreshed every morning; you don't build them yourself. Metric definitions live in one place, so chat and dashboards give the same number. A custom dashboard is set up with you by Luca's team.
Is this the same as analysing customer chats or support calls?
No. Luca answers questions about your own connected business data: orders, products, ad spend, traffic and the books. It does not read customer conversations, support tickets or call recordings, so it is not a tool for contact-centre or chatbot analysis.
How many example questions can I start from?
About 50. Luca offers a library of about 50 expert questions you can pin, plus question cards matched to the platforms you have connected. Pin the ones you ask every week, and the cards change as you connect more sources.
What do I need in place before my first question?
A Luca workspace with at least one source connected. Most sources connect by approving read-only access on the platform's own consent screen, and setting up Luca walks through it. Luca reads from every connected system and writes to none of them, as read-only access explains.
