Free Market Research Chatbot Template: Test Price & Brand

The Market Research template is a full consumer research interview delivered as a chat. Instead of emailing a panel a twenty-question grid, you run the same study as a sequence of short exchanges: one question on screen at a time, buttons where buttons make sense, free text where you genuinely want the respondent's own words. The published template ships as a single flow of twenty-nine nodes covering category usage, unaided brand recall, brand personality, concept appeal, purchase drivers, price sensitivity, and a demographic block, finishing with an optional email capture for people who want to hear when the product launches.

The template is populated with a worked example so you can see how a real study is structured rather than staring at empty placeholders. The example brand sells beachwear, the collection is called Origama, and the concept under test is a product called the Sun Seat. Every one of those names is a text field in the builder. Swapping "beachwear" for your own category and "Sun Seat" for your own concept is a find-and-replace exercise, not a rebuild - the question architecture underneath is category-agnostic and is the part worth keeping.

Who Runs This

Brand and insights teams use it for tracking studies, where the same battery of questions is asked every quarter and the value is in the trend rather than any single wave. Product and marketing teams use it for concept testing before a launch, to find out whether a new product reads as appealing and what specifically puts people off it. Founders and small teams without a research budget use it as a way to run a credible study on their own site traffic, because the alternative - a commissioned panel study - is usually out of reach. Agencies use it as a reusable shell they clone per client.

Because it is built in Conferbot's no-code chatbot builder, you edit the questions by clicking a node and typing. You can browse the rest of the survey template library if you need a shorter instrument; this one is deliberately long-form.

The flow runs in five acts: warm-up and category usage, brand perception, concept reaction, pricing, then demographics. Here is the real sequence as published.

Act One: Warm-Up and Category Usage

The welcome node opens with "Hi there!" and hands straight to a single-button node that sets the frame: "And thanks for taking time to tell us more about you and your beachwear preferences!" with one option, "Not a problem!". A second button node follows - "Let's start by finding out your general beachwear usage!" with "Let's do it" - which exists purely to segment the interview into visible chapters. Those two taps cost the respondent nothing and measurably change how long a chat survey feels.

The first real question is a five-point rating: "How often do you use beachwear products?", stored in a variable named often_products. Then the most valuable question in the whole study, an open text box: "What beachwear brands, if any, can you think of?" into the variable brands. That is unaided brand awareness. It is asked before your brand is ever mentioned, which is the only way the answer means anything - once the respondent has seen your name, recall is contaminated. Act one closes with a single-select: "When did you last buy a beachwear product?" offering "< 1 month ago", "Between 1 & 6 months ago", "6 months - 1 year ago" and "More than 1 year ago", into last_buy. That is your recency screen and the variable you will most often use to cut every other answer.

Act Two: Brand Perception

A button node - "Now to tell us what you really think about us..." with "Sure" - signals the shift from category to brand. Then a projective single-select: "How would you best describe our Origama collection if it were a person?" with the options "Down-to-earth", "Cheerful", "Honest", "Charming" and "Other", into describe_collection. The personification framing is a standard brand-personality technique: people who cannot articulate what a brand stands for will happily tell you what kind of person it would be, and the answer maps cleanly onto positioning work.

Act Three: Concept Reaction

A five-point rating asks "How interested are you in the product on a scale of one to five?" into rating_choice. All five rating outcomes route to the same next node, so nobody is screened out for a low score - low scorers are exactly who you want to hear from next. Two open questions follow and they are deliberately a pair: "What appeals you the most about the Sun Seat?" into appeal_sun_seat, then "And what puts you off it?" into puts_off. Asking for the objection explicitly, in its own question, is what stops a concept test collecting nothing but politeness.

The bot acknowledges with "Thank you!" and asks the driver question as a single-select: "Which of the following would influence your decision the most if you were to buy a Sun Seat?" with "Brand", "Quality", "Value" and "Price", into decision_influence. Then a progress message - "You're doing great, we're about halfway through now." - which is the flow's only completion-rate device and worth keeping.

Act Four: Pricing

A button node, "Let's find out what you think about the Sun Seat's pricing..." with "Alrighty!", introduces the price block. A five-point rating comes first: "How do you feel about the Sun Seat price its against competitors?" into price_opinion. Then three open numeric-style questions, each phrased in USD:

• "At what USD price would you begin to think the Sun Seat to be a bargain and a great buy for the money?" into price_you_buy
• "OK. At what USD price would you begin to think the Sun Seat to be so cheap that you would be suspicious of its quality and not consider buying it?" into cheap_price
• "At what USD price would you begin to think the Sun Seat is getting expensive, but still not out of the question?" into too_expensive_price

Anyone who has run pricing research will recognise three of the four Van Westendorp price sensitivity questions: too cheap, bargain, and getting expensive. The fourth - the point at which the product is so expensive it is out of the question - is not in the published flow. Adding it is the single highest-value edit you can make to this template, and it is one duplicated node.

Act Five: Demographics and Opt-In

Two messages set up the classification block: "Let's speak a bit about you!" followed by "The information you provide will remain confidential and can not be used to identify you." Putting the confidentiality statement immediately before the demographic questions, rather than at the top of the survey where nobody reads it, is why this block gets answered.

Then, in order: age range as a single-select ("< 18 years old", "19 - 24", "25 - 34", "25 - 34", "50+") into age_range; "What's your yearly income?" with "< $10k", "$10K - $25k", "$25k - $50k", "$50k - $75k" and "Prefer not to answer" into yearly_income; a location question, "In which country were you born?", into born_location; employment status ("Employed full-time", "Employed part-time", "Self-employed", "Student", "Unemployed") into employment_status; and education ("Less than high school", "High School", "University", "Undergraduate", "Masters") into highest_education.

Note the age list as published: "25 - 34" appears twice and there is no bucket covering 35 to 49. Fix that before you launch or a third of your sample will have nowhere to click. The income list ending in "Prefer not to answer" is the right pattern and worth copying onto the other sensitive questions.

The flow closes with a validated email node - "If you're interested in the Sun Seat, enter your email address and we'll let you know when it makes it to market!", rejecting bad input with "Please enter a valid email address" and storing contact_email data in email - then a final message, "Thanks a lot for your patience!", and an image node where the published template shows a closing visual. The email question earns its place because it is asked last, after the respondent has spent several minutes thinking about your product, and because it is framed as a benefit rather than a data grab. There is no human handover node in this flow; it is a self-service instrument from end to end.

The template uses six different node types, and the choice of type per question is not cosmetic - it determines whether the answer is analysable.

Every stored variable is available downstream in the builder, so you can reference an earlier answer inside a later question's text, branch on it, or pass it to an integration. The chatbot analytics view gives you drop-off per node, which on a study this long is the number you will watch most closely.

Plan on an afternoon for the first wave and about twenty minutes for every wave after that.

1. Open the template and rename the study. Import Market Research from the surveys category. It opens in the drag-and-drop editor as a single block of twenty-nine nodes laid out left to right in conversation order.

2. Replace the worked example. Work through the nodes in order and change three things: the category word ("beachwear"), the collection name ("Origama") and the concept name ("Sun Seat"). Keep the question structure. Keep the order especially - unaided brand recall has to stay in front of every node that names your brand, or the answer is worthless.

3. Fix the two known list problems. Correct the duplicated "25 - 34" age bucket and add a 35 to 49 band. Add the missing fourth Van Westendorp question if you want a full price sensitivity model.

4. Decide where the answers go. Conferbot ships a Google Sheets node and a webhook node. Dropping a Google Sheets node after the email question writes every variable in the flow as a row per respondent, which is all most teams need for analysis in a spreadsheet. If you already have a research stack, a webhook posts the same payload to it; webhooks and API access require Starter or above. See the integrations hub for the full list.

5. Connect a channel and test. Publish to your site with the website widget or to a messaging channel, then run the flow yourself end to end at least twice - once giving high scores, once low - and check that every variable arrives in your sheet with the value you typed.

6. Publish and watch the first fifty responses. Open analytics and look for the node where people leave. On a study of this length the drop-off is almost always in the demographic block or the price block; the fix is usually a progress message or one fewer question, not a redesign.

The free plan covers 600 conversations on one chatbot, which is enough to complete a first wave. Pricing lists the higher conversation allowances if you are running continuous tracking.

Add a screener at the top. The published flow interviews everyone who starts it. Most studies want a qualifying question first - category usage in the last twelve months, age, or market - with a condition node that thanks non-qualifiers and ends the conversation. That keeps your sample clean and your conversation allowance spent on people who count.

Branch the concept questions on the interest rating. Right now all five rating outcomes go to the same pair of open questions. A more informative design sends scores of four and five to "What would make you buy this today?" and scores of one and two to "What would have to change for you to consider it?". The rating node exposes a separate outgoing path per score, so this is wiring, not code.

Rotate the brand personality options. Fixed option order introduces order bias on a list like "Down-to-earth / Cheerful / Honest / Charming". If your study is a tracker, at minimum keep the order identical between waves so the bias is constant.

Make the price questions numeric. The three price nodes are open text, so respondents will type "about 40 bucks" and "$39.99". Swapping them for number-input nodes forces a clean numeric answer and saves you a data-cleaning pass.

Split the demographic block by market. Income bands in USD do not travel. If you run the same study in more than one country, add a condition node on the country answer and route to a localised income list.

Do something with the email. A Gmail node after the email capture can send an immediate confirmation, and the address can be pushed to HubSpot or Zoho CRM as a launch-list contact. If you would rather let interested respondents talk to someone, a live chat handover node can be added after the opt-in - live handover and the unified inbox are on every plan, including Free.

Reuse the shell. The five-act structure is the reusable asset. Teams commonly clone this flow into a shorter post-launch version and pair it with the NPS survey for existing customers and the product-market fit survey once the product is live. For open-ended comment analysis at scale, the sentiment analysis template is the natural companion.

Every Conferbot channel is available on every plan, including Free, so the choice is about where your sample is rather than what you are paying.

The website widget is the default for this template. Intercepting real category traffic - people already on a product page - gives you respondents with genuine context, and it is the only channel where you can trigger the survey on a specific page. For a concept test, that page targeting is the difference between a relevant sample and a random one.

WhatsApp is the right choice when you are re-contacting a known list, such as past buyers or a panel you recruited yourself, and it is the strongest option in markets where WhatsApp is the default messaging app. A twenty-nine node study is long for a messaging thread, but WhatsApp tolerates it better than email does because respondents can answer in several sittings without losing their place.

Telegram works well for community-recruited samples, and Instagram and Messenger put the study in front of a consumer audience you are already advertising to - useful for a beachwear or lifestyle concept specifically, where the follower base is the target market. Slack and Microsoft Teams are the internal-audience option, for testing a concept with staff or a retail partner before it goes public.

The same flow runs unchanged on all of them - see omnichannel for how one bot serves several channels at once. Running one wave on the website widget and another on WhatsApp is legitimate, but note the channel in your results, because channel mix is itself a source of difference between waves.