Three Questions That Start the Conversation

An empty chat box asks the customer to do the work. They have to guess what information the advisor holds, find the right words and decide whether the effort is worth it. Many simply scroll past.

On a product page, three short suggested questions change that. They show what the advisor can answer, and they put the customer’s own doubt into words before the customer has to.

Why an empty box stays empty

A generic invitation such as “How can I help you?” gives the customer nothing to start from. It does not say what information the advisor holds on this product, or whether it holds any.

A specific question does. “Does this bag fit a 16-inch laptop with its sleeve?” tells the customer that the advisor knows the inside dimensions of the bag. One click and the doubt that was holding the purchase back is on the table.

Where good starter questions come from

The best starter questions are not written in a meeting. They come from what customers already ask:

• Search logs, where customers type full questions, as described in Your Search Bar Is Where Customers Ask Questions.
• Customer service emails and chats.
• Reviews, where customers explain what surprised them.
• Past conversations with the advisor.

The questions that decide a purchase tend to repeat within a category. The Skincare Questions a Product Page Rarely Answers shows how much they have in common from one product to the next.

A good starter question should be:

• Specific to the page, not to the store in general.
• Answerable from approved knowledge. Never suggest a question the catalog cannot answer.
• Short, so it reads at a glance.
• Written in the customer’s language and in the store’s tone.

Three is enough

Three questions cover the main kinds of doubt without crowding the page. A useful mix is one question on suitability (“Is it suitable for…?”), one on use (“How do I…?”) and one on comparison (“What is the difference with…?”).

Measure clicks, rotate, keep what works

Starter questions should be treated as hypotheses:

• Count the clicks on each question, per product and per language.
• Rotate the questions that are rarely chosen and test new ones from the latest customer questions.
• Fix the knowledge when a question is clicked often but its answer is weak. That is a gap in the catalog, and Merchant Console is where it gets filled.

Clicks measure interest, not effect. Whether the conversations change purchases is a separate question, answered by comparing visitors who can use the advisor with a holdout group. Attribution vs Incrementality in AI Commerce explains the difference.

How Brand Agent does it

On each product page, Brand Agent shows three starter questions drawn from customer requests about that product and its category, in the customer’s language. Every click is counted. Questions that are rarely chosen are rotated out for new ones, and the answer behind each question rests on knowledge the merchant has approved.

FAQ

Can the merchant propose its own questions? Yes. They are counted and rotated like the others, so the data decides which ones stay.

Do starter questions replace the search bar? No. They work on the product page, where the doubt is about one product. The search bar receives questions before the customer has found the right product.

How many products need starter questions? Start with the products that receive the most questions or the most traffic, then extend as the knowledge grows.

Want to see which questions your customers would click? Request a demo.