Guided selling: how to recommend products in chat and quizzes

In a good shop, nobody hands you a spreadsheet of forty products. Someone asks what it's for, who it's for and what you've tried before, then points at two options and tells you why. Most online stores skip that conversation. They show filters, specs and a grid, and leave the shopper to work it out.

Some shoppers enjoy that. Many give up, or buy the wrong thing and send it back. Guided selling puts the conversation back. Here's how to set it up in chat, in quizzes and with AI, and how to know which one your store needs.

Why shoppers who can't choose don't buy

Choice problems rarely show up as "choice problems" in your reports. They look like this:

• Long browsing, no cart. Shoppers view many product pages in one category and leave.
• The same chat question every day: "What's the difference between these two?", "Which one is best for…?".
• Returns marked "not as expected" or "wrong size". The product was fine; the choice was wrong.
• Searches that find nothing, because shoppers describe a need ("gift for a runner") and your catalogue is organised by product type.
• High traffic on comparison pages that nobody turns into a purchase.

Each of these is a decision you could help with.

The formats of guided selling

There is no single right tool. Match the format to the decision.

Format	Best for	Effort to set up	Weak spot
Product finder flow in chat	Medium catalogues with a few clear questions (size, use, budget)	Low	Rigid; can't handle unusual questions
Recommendation quiz	Structured choices with a clear right answer (skin type, taste, fit)	Medium	Extra clicks; shoppers may skip it
AI agent in chat	Open questions in the shopper's own words	Low to medium	Only as good as your product data
Expert live chat	Expensive, technical or custom purchases	Ongoing staff time	Limited to your opening hours
Personalised recommendation widgets	Large catalogues with lots of traffic	Medium to high	Doesn't explain why a product fits

Most stores combine two: a quiz or product finder for the main decision, and chat (AI plus people) for everything the quiz didn't cover.

Step 1: Find the decisions that need help

Look for the moments where shoppers hesitate:

1. Read chat transcripts and emails for "which", "difference", "best for", "recommend" and "between".
2. Check site search for searches with no or poor results, and for needs-based terms such as "for beginners" or "for small spaces".
3. Read return reasons. "Too small", "not what I expected" and "wrong model" point to a decision you could guide.
4. Ask your best salesperson what they ask customers first. That's your first quiz question.

Pick one category to start, usually the one with the most questions or the most returns.

Step 2: Write the questions an expert would ask

Good guided selling asks about the shopper, not about the product:

• Use before specs. "What will you mostly use it for?" beats "How many watts do you need?".
• Three to five questions. Every question should change the recommendation.
• Plain answers. "I run a few times a week" instead of "Intermediate".
• An "I'm not sure" option that leads somewhere sensible, often a person.

Example for a coffee store:

1. How do you make coffee at home? (espresso machine, filter, cafetière, not sure)
2. How do you like it? (bright and fruity, balanced, rich and chocolatey)
3. Milk or no milk?

Three questions, and each one narrows the beans.

Step 3: Map answers to products, and clean your data

Recommendations are only as good as the product data behind them:

• Tag products with the attributes your questions use: use case, level, roast, skin type, size range.
• Write descriptions that say who the product is for, not just what it is. AI agents read them.
• Keep stock in sync, so you never recommend something that's sold out.
• Choose a sensible default for every path, so no shopper ends up with "no products found".

Step 4: Choose where it appears

• Category pages: a "Help me choose" entry point above the product grid.
• Product pages: a chat prompt for shoppers comparing similar items. See our guide to proactive chat messages for timings.
• Homepage and ads: a quiz as a landing page for broad campaigns ("Find your perfect…").
• Chat at any time: an AI agent that can recommend products when a shopper simply asks.

Step 5: Hand over to a person for big decisions

Automation should narrow the choice, not close the door. Offer a person when:

• the cart value is high or the product is technical;
• the shopper picks "I'm not sure" twice;
• the question is about compatibility, custom work or bulk orders.

The person should see the quiz answers or chat history, so the shopper never starts again.

Step 6: Use the answers in your marketing

Every answer is first-party data, freely given. Use it with consent:

• Send results by email with the recommended products and a short explanation.
• Segment your list by answers: espresso drinkers get espresso launches.
• Spot catalogue gaps: many shoppers asking for something you don't sell is a product idea.

For collecting emails alongside recommendations, see how to grow your store's email list.

Step 7: Measure whether it works

Metric	What it tells you
Start and completion rate	Whether shoppers want help and the questions are easy
Conversion after a recommendation	Whether the recommendations are right
Average order value	Whether guidance leads to better-matched or bigger baskets
Return rate for guided orders	Whether shoppers chose better
"None of these" or handover rate	Where your questions or catalogue fall short

Common mistakes

• Too many questions. Every extra question loses shoppers.
• Forcing an email before results. Show the answer first, then offer to send it.
• Recommending everything. Two or three products with a reason beat ten "matches".
• Spec-based questions that only experts can answer.
• Set and forget. New products and discontinued lines break quiz paths.

Start with one category

1. Pick the category with the most "which one?" questions or returns.
2. Write three to five needs-based questions with your best salesperson.
3. Tag the products in that category with the attributes the questions use.
4. Launch it in chat as a product finder Flow or through your AI agent, with a person as the fallback.
5. After a month, compare conversion and returns for guided and unguided shoppers, then decide whether a dedicated quiz is worth adding.

Guided selling works best alongside good automation. See the chatbot flows every online store should set up, or compare tools in the live chat category.