AI recommendations with co-brand personalization

What it does

When workspaces are grouped as co-brand workspaces, recommendation campaigns in any workspace of the group can return items from across all grouped workspaces. The recommendation model is trained on behavioral data — item views, clicks, and purchases — from all workspaces in the group. This improves personalization for all customers, not only those whose profiles exist in multiple workspaces:

• Customers who exist in multiple workspaces receive recommendations that draw from the combined item catalogs of all brands.
• Customers who exist in only one workspace also benefit, because the model learns cross-brand behavioral patterns from other customers who shop across brands.

The model handles catalog differences gracefully. Differences in feed schema, such as one workspace using a brand field and another using a producer field for the same concept, are handled automatically. Adding a workspace to the group only adds training data — it does not remove or replace data from any other workspace in the group. This does not guarantee that recommendation quality will match or exceed that of a single-brand model, though: if brands sell largely unrelated product types, for example footwear and cosmetics, prediction quality can vary.

Example: Workspace A is a sports brand, workspace B is a casual clothing brand. A recommendation campaign in workspace B can return sports clothing or accessories — and vice versa, a campaign in workspace A may return casual clothing items.

Recommendation models

Co-brand personalization works for all recommendation models.

Modes

Co-brand workspaces operate in two modes for AI recommendations. The mode is determined automatically based on whether a common profile identifier is configured for the workspace group, and cannot be changed from the UI:

• Events — used when no common identifier is configured. The AI model is trained on events from all workspaces (item views, clicks, purchases), but customer profiles are not unified across workspaces. Recommendations are still improved compared to a single-workspace model, because the model learns from cross-brand behavioral patterns.
• Model & events (Full) — used when a common identifier is configured. In addition to combining events, the AI model unifies customer profiles across workspaces, enabling identity-aware personalization and delivering the best recommendation quality.

Requirements

See General requirements. For AI recommendations specifically:

• A common profile identifier is not required. If one is configured for the workspace group, it operates in Model & events (Full) mode; if not, it operates in Events mode. See Modes.

Setting up

1. Create a group of workspaces.
2. Contact the Synerise support to enable co-brand personalization in AI recommendations. The support team will confirm whether your workspace group meets the requirements and determine the appropriate mode.
3. Once enabled, proceed to creating an AI recommendation.
4. During the configuration, proceed to Additional settings.
5. Enable the Co-Brand Personalization toggle.
   • The badge next to the toggle label indicates the mode configured for your group.
   • The filter applied to the recommendations will consider items from co-brand workspaces.
6. You can preview the recommendation results.
   You can enable or disable inclusion of items from co-brand workspaces in the preview to compare results between the state of the option.

Previewing results

Instructions on previewing recommendation results for AI recommendations with co-brand personalization enabled are available in "Previewing recommendations".