AI Recommendations and Best sellers [OCU]
Learn how OCU's smart AI Funnel functionality can automatically build an optimized funnel based on customer order history!
Jeff Maxfield avatar
Written by Jeff Maxfield
Updated over a week ago

AI Recommendations and Best sellers is a limited release. If you'd like to be added to this new features and provide your feedback, please contact our support team.

Overview

When you launch the OCU app for the first time (when AI functionality is available), the AI algorithm starts collecting order history data and processes it to generate OCU Recommended Products and Best Sellers lists. Based on this data, a smart AI Funnel is auto-generated!

On the Dashboard (Homepage) of the OCU app you will see the AI learning progress bar running:

The progress bar displays 3 steps of the AI algorithm workflow:

  • Collecting order history

  • Processing order history

  • Training AI model

Once the workflow is completed, the AI Funnel is auto-generated and a success message shows. Click the "View Funnel" button to review the generated funnel:

The AI learning progress bar can be closed if desired by clicking the X icon in the top right corner. It will not be displayed again until the app is re-installed at which time the AI training process restarts automatically.

In some rare cases when the AI model is trained successfully, the smart AI Funnel is unable to be created. In this case you can still use OCU Recommended Products and Best Sellers functionality, however the funnel will need to be created manually by clicking the "Create Funnel" button:

If your store does not have enough order history data for AI model training, an error state is shown. OCU will attempt to train the algorithm as more orders are processed on your store. You can check the progress status at Settings > AI Algorithm:

AI Funnel

A default AI Funnel is created after training of the AI algorithm model, when we can receive at least 1 best selling product and at least 1 recommended product for that best selling product.

The default funnel is created with an “Unpublished” status and displayed on the Funnel list page along with manually created funnels. It will be named “{AI Funnel}” and have a ‘?’ icon beside its name in the list. By hovering your cursor over the icon the hint text will be displayed:

The default AI Funnel has the following structure:

Triggers condition

Best selling products

Trigger operator

Is any

Trigger product

Maximum number of generated Best selling products limited to Top-5 (from 1 to 5 bestselling products, apply maximum possible, but not more than 5 products)

Post-Purchase Upsell Offer #1

OCU Recommended product with the highest score (generated based on the triggers setup)

The default AI Funnel is created only once when conditions become appropriate for its creation. It can be done from the first start of AI model training or after its retraining. *If the funnel is deleted, it will not be created again.

You will be able to manage the default funnel the same way as for manually created funnels, however A/B testing is not available for the OCU recommended products.

Using "Best Sellers" as a Trigger

A new trigger type "Best selling products" is added to the list of trigger conditions. Select this option to set your best selling products as a trigger for the funnel.

Available operators for this condition are is any / is not. By clicking the "Add Product(s)" button, your top 10 best selling products list will be displayed.

Check the box next to each one you'd like to use as a trigger for the funnel, then click the "Add" button:

📝For the best sellers, it displays the list of products generated by the AI model and sorted based on the product score (selling rate) from highest to lowest. The list is limited to a max of 10 products, but can contain less than 10 products (for small or newly registered stores without a rich order history). If no best selling product(s) are available for the store, then the trigger condition “Best selling products” will not be available.

OCU will check the availability of best selling product in the store and in case it is out of stock/hidden or deleted, it will not be displayed in the list of products for selection.

If the product becomes unavailable after funnel has already been created, then the funnel will become “Unpublished” automatically and you will see an error tooltip on the funnel details page:

Create “OCU Recommended products” product(s) and “Bestselling product(s)” offer

OCU Recommended Offers will be generated based on the Trigger product(s) used for the funnel, so below are some recommendations:

  1. For using an OCU Recommended Offer, you need to set up at least 1 "product" or "product variant" based trigger.

  2. We recommend using a maximum of 5 trigger products for better recommendation results. The fewer triggers the funnel has, the better the AI recommended products list will be received as a result.

  3. When the bound “trigger product -> related OCU Recommended offer” is set up in the funnel, if you change the trigger product then the OCU Recommended offer can become irrelevant.

When the AI model training is completed successfully, then the new option “Offer your best selling products" becomes available in the “Create offer” popup.

When the AI model training is completed successfully AND you've already set a Product or Product Variant based triggers in the funnel, then the option “Use OCU Recommended products” becomes available in the “Create offer” popup as well:

The “Use OCU Recommended products” option will not be available when:

  1. Trigger does not contain product and/or variant-based triggers.

  2. ALL products added as a trigger to the funnel do not contain purchase history, so recommendations can't be generated by the AI model.

Click on the “Continue” button and the “Add product” popup is opened (for Best sellers or OCU Recommended products respectively).

Add Best seller / OCU recommended products as an offer

There are 2 new popups implemented, for adding a Best selling product(s) and adding OCU Recommended product(s):

Add OCU Recommended product popup:

  1. Match score value converted to the percentage value displayed (1). Reflects how relevant recommended products are, based on the AI model analysis.

  2. You can select up to 3 products from the pop-up on post purchase, multi-selection is available (2).

  3. When max (3 for post-purchase) number of products are selected, multi-selection becomes unavailable (3)

  4. When the total amount of active trigger products and unique product variants in the funnel is >5, a notification will be shown in the popup (4)

  5. Loader state of the popup (5)

OCU considers trigger conditions and operators in the funnel to calculate the number of active triggers. If OR operator is used, then we count products for every condition separately, if AND is used - we count products total.

Add Bestselling product popup:

  1. TOP-3 product’s rating is displayed (1).

  2. You can select up to 3 products from the pop-up on post purchase, multi-selection is available (2). If 1 product is selected, then the OCU system inherits “single product” behavior of offer displaying on the funnel details page. If multiple products are selected, then the OCU system inherits “multiple product” behavior of offer displaying on the funnel details page.

  3. When max (3 for post-purchase) number of products are selected, multi-selection becomes not available (3)

  4. Loader state of the popup (4)

When at least 1 product is selected, the “Add product” button becomes active and can be clicked to confirm adding product(s) as an offer. A/B testing is not available for the OCU recommended products.

AI Model retraining

The AI Model retraining process is supported by OCU automatically on a regular basis or can be initiated manually from the Settings > AI Algorithm section.

During the retraining (auto or manually initiated) you will not see the AI learning progress bar on the Dashboard (Homepage), but will be able to see all necessary info on the AI settings page.

After the retraining, the AI model returns the lists of new Best Sellers and updates Recommended products. The last update date is available on the AI setting page as well:

📝 From time to time, funnels with best sellers and OCU recommended products should be updated manually in order to keep the setup of bestselling products and recommendations in the funnels up to date. OCU does not track these updates dynamically after retraining of the model. So after the retraining, the “best practice” recommendation is to go to the funnels and check best sellers and recommended products and replace them with the updated ones from the newly generated lists.

Auto-retraining

When the AI model is trained for the first time (successfully or not), OCU starts to track new order history records per store. As soon as the number of newly created orders is reached, the AI model retraining starts automatically.

The number of new order history records will be unique per each store. So based on the logic below, OCU defines this number of order history records per store.

OCU System should define order history records number by logic:

  • Collect order history records for 3 months period from current date per store

  • Calculate the average number of order records for 3 months period per store

  • IF the calculate average number of order records is ≥ 100 records THEN set calculated average order number value as a condition for AI Algorithm re-training

  • IF the calculate average number of order records is < 100 records THEN set minimum order number value = 100 as a condition for AI Algorithm re-training

  • Keep applied order number value until the next re-training start

📝 If there were no funnels with OCU recommended or Best selling products in the store, then after the first launch retraining will be started automatically every 60 days by default.

Manual retraining

Manual retraining of the model can be done from the AI settings page at any time, by clicking the "Update Now" button there:

Not available when retraining is already in progress. The number of new history records for the next auto-retraining will also be defined when retraining is started manually.

Did this answer your question?