A product recommendation quiz that narrows your range to one or two picks
When a shop sells more than a handful of similar things (scents, teas, skincare, bikes), customers stall at the choice. A product finder quiz asks what they care about and points to one or two products with a reason for each. Anticode builds the quiz as an app and reads your products from a collection you edit in the dashboard. Its Maison Ardent template, a perfume house with a scent finder, is a useful example to look at first.
How matching works
Behind every product finder is a simple scoring model. Each product gets attributes (for a scent: fresh, woody, floral, sweet, intensity 1 to 5, season). Each answer adds points to the attributes it suggests. At the end, products are ranked by how well their attributes match the points, and the top one or two are shown. There's no need for anything cleverer: a transparent model you can reason about beats an opaque one when a customer asks "why this one?"
| Question | Answer | Adds to |
|---|---|---|
| When do you drink tea? | First thing | Strong +2, caffeine +2 |
| When do you drink tea? | Evening | Caffeine-free +3 |
| Pick a flavour you like | Citrus | Bright +2 |
| Pick a flavour you like | Toasted bread | Roasted +2 |
| Milk? | Always | Strong +1, roasted +1 |
Questions customers can answer
- Ask about the customer, not the product. "Which of these mornings sounds like you?" works better than "Do you prefer high or low tannin?" for people who don't know the vocabulary.
- Five to seven questions. Fewer feels arbitrary; more feels like a form.
- Picture-style answers for taste and mood, drawn in code as swatches, icons or small illustrations.
- One budget or use question, if price or occasion rules products in or out. Treat it as a filter rather than points.
- A gift path. If people buy for others, start by asking "for you or for someone else?" and rephrase questions accordingly.
The result screen
Show one main pick and one alternative, each with a sentence that ties back to the answers ("You chose evening and citrus, so this caffeine-free lemon verbena fits"). Add the price and a link to the product's own page or checkout. Let people retake the quiz or see the full range. If no product scores well, say so and show your two most popular ones rather than forcing a weak match.
Products from your own catalogue
The part of this that changes most often is the product list, so it shouldn't live in the quiz's code. Anticode apps can read owner-managed content from the project's CMS collections. Describe it in the request, for example: "A tea finder. Products are a collection with name, description, price, buy link, and attributes strong, bright, roasted, floral and caffeine-free as numbers." The build declares that collection with realistic starting entries, which then become yours to edit in the dashboard.
When you add a tea, change a price or retire a flavour in the dashboard, the published quiz picks it up and scores the new list. Inside the app the collection is read-only: visitors only ever see it, and the quiz has no editor of its own. That separation is what lets someone who isn't technical keep the finder current.
Seeing it done: Maison Ardent
The template gallery includes Maison Ardent, a perfume house with a scent finder, written by Anticode's model. Using a template copies it into your project with no model call and nothing off your limits. It's a good way to see the question flow and result design before describing your own.
Buying and following up
The app can't take payment or add to a real basket, because Anticode apps have no network and no payments. Each product's link, such as its page on your own shop, opens in a new tab. Anticode's payment link is one per account, used by every buy button, so for a range of products link to product pages or a product list at your provider. If you'd rather talk to undecided customers, add an optional "email me my match" form: it reaches you through Anticode's form pipeline, and you reply in person.
Limits
- No stock levels. The quiz recommends whatever is in the collection; remove sold-out products there.
- No purchase data. It can't learn from what people bought, because it doesn't see sales.
- No product photos from your shop. Visuals in the app are drawn in code, not loaded from your store.
- A curated range, not a warehouse. The app receives a read-only copy of the collection inside the page, which suits dozens of products rather than thousands of SKUs.
Questions people ask
Can a product recommendation quiz work for a big catalogue?
It works best for a curated range. For hundreds of products, ask the quiz to recommend a category, then link to that category on your shop.
Do I see which products people were matched with?
Only if they submit a form. Otherwise their answers and result stay in their browser.
Can I change the questions later?
Yes, ask for the change in words. The questions live in the app; the products live in the collection.