Aiden's recommendations algorithm is based on scoring and weighing each individual product in your catalog. The score of each product is determined by you when you link the products to the responses.

Each product should be scored as a bad, neutral, good or perfect match for each answer in the conversation:

In the advice, the customer will mainly see (available) products that have a perfect or good match to their answers.

A product that matches poorly with one or more of the chosen answers will probably not be shown in the advice, or only very occasionally.

Would you like to be able to completely exclude products with poor matches to a particular question from the advice? For example, mattresses that are only suitable for children and therefore do not need to be shown in an adult mattress recommendation? You can exclude these products in Advice Settings in the Matching section.

‍

'Overview' is your dashboard where you see the most recent status of your matching. On the left is a 'to do list' with the questions and products that still need to be matched. The counter at the top gives an indication of how 'product rich' your product finder is.

The Start matching button leads to the Match tab, where you can match products to your answers one by one.

Would you rather get started with multiple unmatched products at once? Just click on the number of unmatched products for any of your answers to match them.

In the overview on the right, we show the distribution of your matching for each question. It shows exactly how many products are unmatched, or a poor, neutral, good or perfect match for each answer:

This quantitative overview helps you understand the overall distribution of your matching across responses. This allows you to see if it matches the expected distribution. For example: as an expert, you may know that your catalog contains at least 3 "on-ear" headphones and 1 "over-ear" - which means these scores look good!

The fastest way to match all products is to use rules. In a rule you determine when a product is a perfect, good, neutral or bad match for an answer. You use product features to set up the rules.

When a new product is added to the catalog, this product will immediately be matched for the answers for which matching rules have been set.

Read more about matching using matching rules here.

‍

Sometimes not all answers to all questions can be matched with rules, fortunately we have a solution for this too. You can match these answers for multiple products at the same time with bulk matching or you can do the matching per product.

You can read more about bulk matching here.

And you can read more about matching per product here.

‍

Matching a numerical question is simple: you define the matching rule once for all products!

1. In the Overview or Bulk match tab, locate the numeric question you want to match:

In the Bulk match tab, use the icon for "Set a rule" to match the numerical question:

‍2. Select a numeric attribute from your dataset to match the numeric query. Don't see the attribute you want? Make sure it contains only numeric values.

3. Define when products are a match. For example, when their value in the price column...

a: equals the user's answer: the value exactly matches the customer's answer.

b: is greater than the user's answer: the value is greater than the given answer.

c: is smaller than the user's answer: the value is smaller than the given answer.

d: is within a range of the user's answer: the value is within a certain range of the given answer.

4. When using step 3d described above, define the range at Range.

5. Click Save changes and you're done setting up your rule!

‍

Matching a drill-down question works exactly the same as any other question. Since the "deepest" level of your drill-down question contains the final answer, you only need to match that level. We do show the full context for each answer, as you can see in this example:

Matching a drill-down question automatically is even easier. Read more: Drill-down question.