> For the complete documentation index, see [llms.txt](https://docs.codegiganten.de/plugin-documentation/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.codegiganten.de/plugin-documentation/en/apps/cross-selling-pro-ai-cart-cross-selling-bundles-and-sets/weighting-and-boosters.md).

# Weighting, diversity and boosters

How weighting, profiles, fine-tuning, diversity and boosters decide the order of recommendations.

Once the filters have decided which products qualify, the weighting determines the order in which they appear. Boosters give individual products extra points, and diversity makes sure the list is not just products of one brand side by side. The defaults suit most shops. Change one setting at a time and check the result in the [preview](https://docs.codegiganten.de/plugin-documentation/en/apps/cross-selling-pro-ai-cart-cross-selling-bundles-and-sets/preview-and-sandbox.md).

## The order of the calculation

For recommendations based on a product (“Similar products”, also in the cart), InsightSearch calculates in this order:

1. **Filters:** only products that pass all filters remain. See [Filters and rule builder](https://docs.codegiganten.de/plugin-documentation/en/apps/cross-selling-pro-ai-cart-cross-selling-bundles-and-sets/filters-and-rule-builder.md).
2. **Scoring:** each product gets a similarity from the nine weights.
3. **Boosters:** extra points are added on top.
4. **Diversity:** in mode “Mixed” the list is made more varied.
5. **Relax instead of empty:** if there are not enough products, category or product type are relaxed, but never stock, rules or exclusions.

{% hint style="warning" %}
**Applies only to recommendations based on a product.** With “Bestsellers”, “Trending now”, “Recently viewed”, “You may also like”, “Goes well with”, “Continue where you left off”, “Buy again” and “Inspired by your wishlist”, the order comes solely from the respective behaviour signal. Weighting, boosters and diversity have no effect there, even if the form shows them. Filters, rules and exclusions still apply.
{% endhint %}

## How the weights work

Weights are not fixed points. InsightSearch divides each weight by the sum of all weights. A slider therefore always acts relative to the others: all at 1 behaves like all at 0.5. The bar “Share of similarity” shows the current distribution in percent.

**Example:** you raise “Price” from 0.2 to 0.6. The price share grows from 12 % to about 29 %, brand and category count correspondingly less.

Content similarity and price are scored relative to the other candidates: the most similar product in the pool gets the top value, the least similar the lowest. If the price is missing on the reference product or the candidate, price does not count for that product.

### Shares that need data

Some shares in the bar are striped. They only count when the required data exists. Otherwise their share is spread across the rest. Below the slider you see whether the requirement is met, for example “✓ Fulfilled: the AI preparation is finished for … % of products.”

| Slider              | Only takes effect once …                                                                                                                                                            |
| ------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| “Content (AI)”      | the AI preparation of the products is finished and “Semantic similarity (AI)” is switched on in the AI section                                                                      |
| “Shopper behaviour” | visitors view or buy products together. Visitor tracking collects this, InsightSearch evaluates it every night. “Include purchase behaviour” must be switched on in the AI section. |
| “Popularity”        | products have already been viewed or bought. The value comes from the insight analysis, which you control in the AI section.                                                        |

In the AI section you can also strengthen or weaken the influence of content similarity. If a factor other than 1 is set there, the bar takes it into account. If an AI feature is switched off only in some sales channels, the form points this out.

### Sliders that have no effect in a mode

In mode “Only same category” all products share the same category, so the “Category” slider does not change the order. The same applies to “Brand” in mode “Only same brand”. The form then shows the note “Has no effect on the order in the selected mode …”.

If the product type is set to “Preferred (same group)”, a fixed share “Product type (preferred)” is added to the bar.

## The profiles at a glance

A profile sets all nine weights at once. The table shows what each profile sets. “Balanced (default)” leaves all fields empty and therefore uses the InsightSearch defaults.

| Slider            | Balanced | Prefer same brand | Same properties | Similar content (AI) | Shopper behaviour first | Same price level |
| ----------------- | -------- | ----------------- | --------------- | -------------------- | ----------------------- | ---------------- |
| Category          | 0.35     | 0.35              | 0.35            | 0.3                  | 0.3                     | 0.35             |
| Brand             | 0.35     | 0.6               | 0.2             | 0.2                  | 0.2                     | 0.2              |
| Content (AI)      | 0.3      | 0.3               | 0.2             | 0.6                  | 0.2                     | 0.2              |
| Price             | 0.2      | 0.2               | 0.15            | 0.15                 | 0.15                    | 0.6              |
| Shopper behaviour | 0.2      | 0.2               | 0.1             | 0.1                  | 0.6                     | 0.1              |
| Popularity        | 0.1      | 0.1               | 0.05            | 0.05                 | 0.3                     | 0.1              |
| Context           | 0.1      | 0.1               | 0.5             | 0.1                  | 0.05                    | 0.1              |
| Attributes        | 0.05     | 0.05              | 0.3             | 0.05                 | 0.05                    | 0.05             |
| Tags              | 0.05     | 0.05              | 0.05            | 0.05                 | 0.05                    | 0.05             |

“Similar content (AI)” needs the finished AI preparation, “Shopper behaviour first” needs enough order and click data. Without it, the profile carries a warning sign with the note “Not useful for your shop yet: …”.

{% hint style="info" %}
“Prefer same brand” is not the same as the mode “Only same brand”: the profile moves the same brand up, other brands remain possible. The mode excludes them.
{% endhint %}

## Fine-tuning and diversity

In expert mode the “Fine-tuning” fields appear below the weighting. These values are not shares, they act in addition.

| Field                       | Effect                                                                                                                                          | InsightSearch default |
| --------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------- | --------------------- |
| “Max. price deviation”      | price band for price similarity. 0.35 means products within ±35 % of the base price count as price-similar, beyond that price no longer counts. | 0.35                  |
| “Prefer same category”      | switch. Only with it does the “Category bonus” take effect.                                                                                     | off                   |
| “Category bonus”            | favours the same category, 0 = no effect, 1 = strong                                                                                            | 0.1                   |
| “Brand penalty”             | deduction for products of a brand already in the list. Increase for more brand variety.                                                         | 0.25                  |
| “Category penalty internal” | deduction for products of a sub-category already in the list. Increase for more variety.                                                        | 0.1                   |
| “Diversity”                 | only in mode “Mixed”. 1 = similarity only, 0 = maximum variety of brand and category.                                                           | 0.75                  |

Diversity fills the places one after another: for each place the similarity counts, minus the penalties for brands and categories already chosen. In the modes “Only same brand” and “Only same category” this step is skipped.

## Boosters

Boosters are business rules: if the condition matches, a product gets extra points, regardless of how similar it is. The points are added on top of the similarity.

### Prefer products in stock

The switch “Prefer products in stock” under “2. Filters” is a booster: products in stock move up, out-of-stock products remain possible. Many presets switch it on. If out-of-stock products should not appear at all, use the building block “Only in stock” in the rule builder.

### Section “4. Boosters”

This section only appears in expert mode.

| Field            | Effect                                                               |
| ---------------- | -------------------------------------------------------------------- |
| “Brand boost”    | strengthens products of the same brand                               |
| “Category boost” | strengthens products of the same category                            |
| “Path boost”     | strengthens products from the same category path (parent categories) |

Values range from 0.1 to 10, the value 1 means no effect. Even if you leave the fields empty, InsightSearch gives the same brand and the same category a small default boost. The extra points are capped, so a booster cannot outweigh similarity without limit.

{% hint style="info" %}
The “Minimum similarity score” checks similarity without boosters. A booster therefore does not lift an unsuitable product above the quality threshold.
{% endhint %}

### What does not exist

* **No custom booster rules:** you cannot create your own rules such as “Same brand +0.1” in the Administration. The text in the form is just an example, you achieve that effect with “Brand boost”.
* **No margin or new-arrival booster:** if you want to show only products of one margin class, transfer a custom field via the [data fields](https://docs.codegiganten.de/plugin-documentation/en/apps/cross-selling-pro-ai-cart-cross-selling-bundles-and-sets/data-fields.md) and filter on it in the rule builder. That is a strict filter. For targeted placements, use “Fixed products” in the template.

Do not confuse: the [Conversion Booster](https://docs.codegiganten.de/plugin-documentation/en/apps/cross-selling-pro-ai-cart-cross-selling-bundles-and-sets/conversion-booster.md) is a separate element for visitors from Google Shopping and has nothing to do with these boosters.

## Typical adjustments

| Goal                                                      | How to do it                                                                                               |
| --------------------------------------------------------- | ---------------------------------------------------------------------------------------------------------- |
| keep recommendations within budget                        | profile “Same price level” or lower “Max. price deviation”                                                 |
| more brand variety                                        | increase “Brand penalty” or lower “Diversity”                                                              |
| accessories instead of alternatives                       | profile “Shopper behaviour first” once there is enough data; plus product type “Matching types” if enabled |
| fashion, furniture, decor with well-maintained properties | profile “Same properties”                                                                                  |
| products in stock first                                   | switch on “Prefer products in stock”                                                                       |
| back to the start                                         | profile “Balanced (default)” and, in the “Expert mode” card, “Reset all expert settings”                   |

Why a product sits in its position is explained in the recommendation sandbox, which breaks down each recommendation by signal. See [Preview and sandbox](https://docs.codegiganten.de/plugin-documentation/en/apps/cross-selling-pro-ai-cart-cross-selling-bundles-and-sets/preview-and-sandbox.md).

## When does a change take effect?

Changes to weights, fine-tuning and boosters apply from the next recommendation after saving. The AI switches in the AI section also take effect immediately. New purchase signals and popularity, on the other hand, are created in the nightly evaluation.

## Read more

{% content-ref url="/pages/bOsKhfEoKG9Mcf9pbuHy" %}
[Filters and rule builder](https://docs.codegiganten.de/plugin-documentation/en/apps/cross-selling-pro-ai-cart-cross-selling-bundles-and-sets/filters-and-rule-builder.md)
{% endcontent-ref %}

{% content-ref url="/pages/aG6vtzB13aErytbtzpSe" %}
[Preview and sandbox](https://docs.codegiganten.de/plugin-documentation/en/apps/cross-selling-pro-ai-cart-cross-selling-bundles-and-sets/preview-and-sandbox.md)
{% endcontent-ref %}

{% content-ref url="/pages/A5tmmG01QN0sNwbBOlS6" %}
[How recommendations work](https://docs.codegiganten.de/plugin-documentation/en/apps/cross-selling-pro-ai-cart-cross-selling-bundles-and-sets/how-recommendations-work.md)
{% endcontent-ref %}

{% content-ref url="/pages/hSRY2MxmbdY5dCCa9KY4" %}
[Creating recommendation templates](https://docs.codegiganten.de/plugin-documentation/en/apps/cross-selling-pro-ai-cart-cross-selling-bundles-and-sets/recommendation-templates.md)
{% endcontent-ref %}


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://docs.codegiganten.de/plugin-documentation/en/apps/cross-selling-pro-ai-cart-cross-selling-bundles-and-sets/weighting-and-boosters.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.
