Skip to main content
All CollectionsOnboarding
Configuring Custom Recommendation Rules
Configuring Custom Recommendation Rules

Learn how to configure your product recommendation surfaces to set up custom rules or campaigns

Axel Larsson avatar
Written by Axel Larsson
Updated over a week ago

Introduction

At Depict, we believe in minimizing manual work for e-commerce teams through automation. However, there are cases where you might want to manually override the recommendations provided by Depict. This article will show you how to do it easily via the Depict app.

When should I make custom recommendation configurations?

In most cases, we recommend you sit back and let Depict's algorithms do the heavy lifting. That said, Depict does not always have the store-specific expertise and context that a human e-commerce manager or merchandiser might have. Some example use cases where you might want to manually override the AI recommendations include:

  • Store-specific campaigns, such as when you want to promote specific products due to a sale period or similar

  • General campaigns, such as holiday campaigns, seasonal drops, etc.

  • Nisched product-specific matching rules - depending on how your product catalog looks, you might prefer not to recommend products of type A with products of type B.

How to configure custom recommendation rules

The process of configuring custom recommendation rules using the Depict app is outlined in the following step-by-step guide:

  1. Go to app.depict.ai and sign in using your Depict account.

  2. In the left sidebar, navigate to the "Recommend" tab.

  3. Click on the "Create new" button to create a new custom recommendation configuration. You may rename it by double-clicking the name at the top of the window.

  4. Choose what recommendation surfaces you want the rule to be applied to.

  5. Choose when you want the recommendation configuration to be active. You can either schedule configurations to run between two dates, inactivate it completely, or set it to run indefinitely (until it is inactivated or deleted).

  6. Click Continue.

  7. For recommendation surfaces that are product dependent (e.g. Related, Cross-sell, Checkout), choose what products and/or categories you want the configuration to apply to. If you want it to apply across all products and categories, leave this section blank.

  8. Choose what products or categories to "Always show" or "Never show" for the surfaces you specified in the third step. Click the "Add" button in the bottom right corner to confirm.

  9. Save your changes by clicking the "Save" button in the top right corner. The configurations should take effect within a few minutes.


FAQ

Q: Can I make configurations on a market-specific level?

A: No. At this point in time, Depict only supports global configuration rules, meaning recommendation rules are applied across all your markets.

Q: What surface name corresponds to what surface on my site?

A: We are working on making these names more clear. However, as a general rule of thumb:

  • Landing page: Shows popular products and personalized recommendations (based on the shopper's on-site interactions), usually placed on your store's landing page.

  • Related: Shows primarily similar products, with a little mix of complementary products, typically placed on the product detail page.

  • Cross-sell/Afterbasket: Shows a mix of complementary and similar products, typically placed on the product detail page or the post-add-to-cart popup.

  • Checkout: Shows cheaper items that complement the customer's cart to boost average order value, typically placed in the mini-cart or on the checkout page.

  • Recently viewed: Shows the products a customer recently has viewed, typically placed on the landing page or towards the bottom of a product detail page.

  • Search: Shows products that are similar and complimentary to the search results. If no search results were found, the surface defaults to show popular products.

Got more questions?

Feel free to reach out to our Customer Success team using the chat function in the Portal, or send an email to [email protected].

Did this answer your question?