Jump to...
Related guides.
Share this guide:
Managing and adjusting cluster settings
Goal: In this article, you’ll learn how to view and modify cluster settings in PowerRelate. This functionality gives you insight into how clustering is generated and allows you to make immediate adjustments when needed.
What are cluster settings?
Cluster settings provide visibility into how PowerRelate groups products and which relationships are shown. These settings define which attributes (features) are used for clustering and how product relationships are displayed. You can tailor these settings to optimize clustering for your specific assortment.
Use cluster settings when you want to:
- Understand how the current clustering was generated.
- Improve clustering by adding or removing features.
- Control which product relationships are shown.
- Trigger reclustering directly, without contacting the support team.
What types of cluster settings are available?
Cluster settings are divided into two categories: clustering settings (which determine how products are grouped) and relation settings (which determine which product relationships are displayed). For clarity, these are organized into separate tabs in the interface.
Clustering settings
Forced feature exclusion: Features listed here will be completely ignored during clustering. This is useful for excluding irrelevant data that could otherwise distort results.
Cluster splitting features (strict): Clusters are split based on these features. Even empty values form their own clusters. If two products differ on any of these features, they will be placed in separate clusters. Empty values will never be grouped with filled values. A common example is “brand”—each brand will always form its own cluster.
Cluster splitting features (soft): Similar to the strict variant, but more lenient: empty values can be grouped with filled ones. If you’re unsure, use the strict option.
Cluster splitting keywords: Products are split into separate clusters if specific keywords appear in their descriptions. Use this only when feature-based splitting isn’t feasible.
Limit cluster size: When enabled, clusters are limited to a maximum of 100 products. Oversized clusters will be automatically split into smaller ones.
Relation settings
Conditional feature groups: From each group, only one feature will be shown in the product relationships. Add redundant features (e.g., “color” and “color code”) to the same group to avoid duplication.
Required per-segment relation features: These features must always be shown in relationships within each segment. Use them to ensure that the key distinguishing attributes of each segment are visible.
Optional per-segment relation features: These features can be shown when necessary—typically when required features alone aren’t enough to uniquely identify each product.
Use only manually entered relation features: When enabled, only the required and optional relation features will be displayed. Other features will still be used internally for clustering, but won’t appear in the output.
Lower-priority relation features: These are fallback features. They’ll only be used if uniqueness cannot otherwise be achieved.
Maximum number of relations: Limits the number of relationships displayed per cluster.
Saving changes
When you modify any settings, a pop-up will appear asking whether you’d like to recluster all unchecked products.
If you enable the “Recluster all unchecked products” toggle, all products that haven’t yet been verified will be reclustered. This is a significant action that may affect any clusters not yet fully approved. To proceed, you’ll need to type “confirm” in the text field.
Note: Reclustering will not affect any relationships that are already live in your webshop.
Changes are applied immediately, but reclustering can take several hours to complete.
