Product matching for advice flows

Matching decides which products can appear, which products score best and which explanation labels visitors see on product cards.

Where to configure this

• Choose the dataset in Advice flows -> open a flow -> Matching.
• Limit the full product group in Matching -> Filters & Sorting.
• Match manual answers in Matching -> Questions -> open an answer.
• Use dataset values as answers in Builder -> open a question -> Answer source.
• Add boosts in Matching -> Boosts.
• Test paths in Matching -> Test.
• Control product card behavior in Settings -> Recommendations.

How matching works

Build matching in this order:

1. Choose the dataset.
2. Limit the product group for the whole Advice flow.
3. Connect answers to product fields.
4. Add boosts for softer preferences.
5. Test realistic answer paths.
6. Publish when the recommendations are right.

This order keeps the setup easier to understand. If the product group is wrong, answer rules and boosts become harder to judge.

Choose the product group

Use Product filter for broad limits that apply to the whole Advice flow.

Good examples:

• only one category or collection;
• only products in stock;
• only products for one use case;
• exclude accessories or replacement parts;
• keep one Advice flow focused on one brand group.

Use Include matching products when only products that match the filter may appear. Use Exclude matching products when a group should stay out.

Check the product count before continuing. If the count is too low, too high or unexpectedly empty, fix the Product filter first.

Match manual answers

Open Matching -> Questions, choose a question and open the answer you want to match.

Use three patterns:

• Hard requirement: the product must match. Use this when a wrong product should not appear.
• Preference: matching products should score higher, but other products may still appear.
• No preference: the answer should not change the product advice.

Examples:

• "Only show waterproof jackets" is a hard requirement.
• "I prefer lightweight jackets" is a preference.
• "No preference" usually needs no rule.

Keep hard requirements limited. Too many strict answers can leave visitors with no products.

Add explanation labels

Use a short USP label when the product card should explain why a product fits. Good labels are short and visitor-friendly, such as Waterproof, Fits small balconies or Made for daily use.

Do not turn every rule into a label. Show only labels that help visitors trust the recommendation.

Use dataset values as answers

Use a dataset attribute when a question has many possible answers, such as model, size, color, material or compatibility.

1. Open the question in the builder.
2. Set Answer source to Dataset attribute.
3. Choose the product field.
4. Use a dropdown when the list is long.
5. Publish after imports when visitors should use updated values.

This is useful when the answer list changes with your catalog. You do not have to maintain every value manually in the flow.

Use boosts

Boosts move better fitting products higher without removing other products.

Use boosts for softer signals:

• popular choice;
• best value;
• preferred color or material;
• stronger match for a use case;
• products you want to highlight when they still fit.

Boosts should improve ranking, not rescue weak rules. If clearly wrong products appear, fix the Product filter or hard requirements first.

Control the result page

Use Settings -> Recommendations to decide how many products visitors see and which labels are shown.

Good defaults:

• show a small number of products first;
• open product links in the current tab, or turn this off when a new tab fits the webshop journey better;
• use review stars only when review rating is mapped;
• show boost labels only when they explain the advice;
• keep non-matching labels on while testing, then hide them if the public card becomes too busy.

Test before publishing

Test with real customer situations.

1. Choose a common visitor type.
2. Complete the flow.
3. Check whether the first products make sense.
4. Open the explanation on product cards.
5. Repeat with a visitor who should get different products.
6. Make strict rules softer when no products remain.
7. Make strict rules tighter when wrong products appear.
8. Adjust boosts when good products appear in the wrong order.

Publish only when the first results are easy to explain.

Keep it manageable

Use clear product fields and a small number of rules per answer. If your team cannot explain why a product wins, simplify the setup before adding more rules.