E-commerce
June 26, 2026
An undecided customer is not necessarily a cold customer. Often, they want to buy, but they do not yet know which product to choose, which model to avoid, which size to take, or what risk they are willing to accept.
The problem arises when the shop responds with more choices, more promotions, and more pop-up windows. The undecided customer needed clear advice, not one last bit of pressure.
This article #36 stands out from content on product recommendations: here, the subject is conversational advice for the hesitant, with useful questions, honest comparisons, and support in making choices that can also say "this product is not the best one for you."
Summary
What is an e-commerce product assistant?
A product assistant is a conversational guide that helps a visitor choose the right item based on their usage, constraints, and level of confidence. It can live in a chat, a side panel, a collection page, or a "buyer's guide" module on a product page.
Shopify describes AI shopping assistants as agents capable of leveraging the catalog, FAQs, policies, and structured data to accurately answer purchasing questions (Shopify, AI personal shopper 2026).
Its mission is not to direct the customer to the most expensive product. Its mission is to reduce uncertainty and make the choice more comfortable.
Why do indecisive customers leave a store?
Indecision rarely comes from a single block. It often arises from an accumulation: too many options, similar product sheets, conflicting reviews, a poorly visible return policy, an uncertain size, or a doubt about the value.
Choice fatigue: the customer compares too many options without knowing which criterion to prioritize
Perceived risk: fear of making the wrong purchase, of a complicated return, or of poor compatibility
Lack of simple language: technical jargon or marketing promises that are difficult to translate into concrete usage
Blurry comparison: two products seem close, but the useful difference is not visible
Fragile trust: reviews, warranty, delivery, or brand are still insufficiently reassuring
Carti sums up the subject well: many stores do not have a traffic problem, but a decision problem. Customers want to be guided, not left alone in front of the entire catalog (Carti, guided selling 2026).
What is the difference compared to a classic product recommendation?
A product recommendation often infers from browsing history, previous purchases, or merchandising rules. A product assistant dialogues: it asks questions, explains its reasoning and can change its mind depending on the response.
Recommendation: useful for showing similar products, bundles, or best-sellers
Product assistant: useful when the customer is hesitating, comparing, or does not know how to formulate the right filter
Quiz: effective on a structured journey of 3 to 7 questions
Advice chat: best when the request is nuanced or unexpected
Shopify points out that recommendation systems are measured via clicks, conversion, and average cart value, but that they depend on clean data and suitable placement (Shopify, AI recommendation systems).
The product assistant complements these modules when the customer needs an advisor, not just a row of items.
Which profiles of undecided individuals should be recognized?
The competitor: they alternate between two models and want the concrete difference
The cautious one: they check returns, warranty, reviews, delivery time, and security
The novice: they are discovering the category and do not know the selection criteria
The rushed one: they are buying for a gift or an urgent delivery
The perfectionist: they want to be sure about the size, the look, or the compatibility
Each profile requires a different tone. The competitor wants a clear table. The cautious one wants proof. The novice wants a simplified explanation. The rushed one wants a reliable delivery time. The perfectionist wants honest limitations.
Do not force the same script on everyone. An initial question like "What is making you hesitate?" or three buttons "compare", "choose size", "check compatibility" is often enough.
What questions can you ask without turning help into a form?
The rule of thumb: three to five questions maximum, each of which must eliminate options or clarify a decisive criterion.
Start with the usage: "For what main need?"
Ask about the constraint: budget, size, skin type, compatibility, deadline.
Identify the level: beginner, regular, expert, or occasional use.
Check the risk: return, warranty, maintenance, delivery.
Suggest a shortlist of two or three options, not ten.
Shopify 2026 quiz guides generally recommend short, mobile-friendly paths with truly discriminating questions and an explained recommendation at the end. The same logic applies to a conversational assistant.
Beauty example: skin type, objective, sensitivity, preferred texture. Sport example: frequency, surface, level of cushioning, budget. Home example: area, usage, style, maintenance constraint.
How do you recommend without pushing to buy?
Credible advice relies on a justified shortlist. A single recommendation can seem suspicious; ten choices recreate the problem.
Response Structure
Reformulate the need: "you are looking above all for..."
Propose the best choice with two verifiable reasons.
Add an alternative if the budget or usage changes.
Mention an honest limitation.
End with a soft choice: view the sheet, compare, ask for a human.
Heeya describes effective guided selling as an exchange that asks a few targeted questions and then recommends two or three products with a clear reason, without turning the help into manipulation (Heeya, guided selling 2026).
Example: "For daily use, model B is more suitable thanks to its warranty and battery life. If you only use it on weekends, model A is sufficient and costs less."
What product data should be prepared before launching the assistant?
A product assistant does not compensate for a vague catalog. It amplifies what you have structured.
Attributes: material, dimensions, weight, compatibility, level, usage
Variants: size, color, pack, availability, and useful differences
Guides: sizing, care, choice, installation, comparison
Social Proof: customer reviews, UGC, certifications, warranty
Policies: return, refund, delivery, exchange, fees
Exclusions: when not to recommend a product
Shopify emphasizes the importance of a structured catalog and a single source of truth for AI assistants. The more precise the data, the more useful the advice can be.
Also add guardrail rules: never recommend an out-of-stock product, signal a technical uncertainty, do not invent an undocumented compatibility.
Where should the assistant be placed on the website?
The right placement depends on when the hesitation occurs.
Collection page: helping to narrow down choices even before opening the product pages
Product page: answering questions about size, compatibility, reviews, returns, comparisons
Cart: reassuring about variants, shipping, packs, or gifts
Guide article: transforming a reader into a more confident product choice
Floating chat: useful if the trigger remains discreet and contextualized
Gorgias, for example, allows you to adapt the selling style between Educational, Moderate, and Promotional. For undecided customers, the educational style is often the best starting point: it asks questions, shares information, and avoids offering discounts too early (Gorgias, selling style).
On mobile, opt for a "Help me choose" button near the CTA rather than a widget that covers the image, price, or variants.
How to link the assistant, FAQ, and product sheets?
The assistant must not become a permanent band-aid. Every repeated question is a signal that the product page or the FAQ needs to be improved.
Question asked 10 times: add a micro-FAQ on the relevant page
Recurring comparison: create an A vs B table
Sizing doubt: highlight the size guide and fit reviews
Trust doubt: place reviews, warranty, and returns closer to the CTA
Compatibility doubt: add a "works with" table
See automated product FAQ, purchase objections and insights from support.
The ideal loop: conversation, tag, product page correction, bot re-indexing, measurement of the decrease in repeated questions.
Which KPIs measure a healthy decision support?
Open rate: how many visitors use the help feature on the PDP or collection page
Completion rate: how many complete the mini-choice journey
Assisted add-to-cart: add-to-cart after a recommendation or comparison
Assisted conversion: purchase within the session or defined attribution window
Return rate: do the recommended products get returned less or more often?
Advice CSAT: did the customer find the assistance helpful?
Unanswered questions: catalog gaps or FAQ to be corrected
The healthiest KPI is not just conversion. It is conversion with a stable or decreasing return rate. If the assistant sells more but generates more returns, it is pushing too hard or giving poor advice.
Compare a pilot category with and without the assistant for 30 days, with comparable traffic. Also look at the time elapsed between opening the assistant and adding to cart: good advice often shortens the decision-making process.
How does Qstomy help undecided customers without overselling?
Qstomy can act as a product assistant connected to Shopify: it understands the page being viewed, asks helpful questions, compares products, addresses objections, and suggests a justified shortlist.
DTC Scenario
A supplement brand receives 2,000 monthly visits on a collection of 12 similar products. Visitors hesitate between sleep, stress, and energy. Qstomy asks three questions, recommends a maximum of two products, and explains why. Pilot objective: +12% assisted cart additions, 20% fewer repeated questions on product pages, and no increase in return rates.
The guardrail is essential: Qstomy does not push a discount by default, does not recommend an out-of-stock product, and escalates to a human if the request involves a sensitive or undocumented constraint.
Discover the AI sales assistant, Shopify integration, human handoff, and request a demo.
Which playbooks should be launched this week?
Playbook 1: spotting indecision
List the five high-traffic, high-time-spent, low-conversion product sheets. Read the associated chats and note the hesitations: size, comparison, trust, usage, price.
Playbook 2: writing three questions
For a pilot category, write three questions that allow a human agent to offer advice. Each answer must reduce the number of options.
Playbook 3: producing a shortlist
For each profile, prepare a maximum of two recommendations: main choice, budget alternative, honest limitation. Add the reason for each proposal.
Playbook 4: measuring without pressure
Track opens, completions, cart additions, conversions, CSAT, and returns. If conversions go up but returns also go up, rework the advice rules.
Useful linking
Assistant vs reco: shopping assistant vs recommendations
Discovery: discovery chatbot vs search
Product questions: chatbot product questions
KPIs: e-commerce chatbot KPIs
Enzo
June 26, 2026
