This page describes the recommendation models provided by Discovery for Media Recommendations, their default serving configs and optimization objectives, their available customizations, and their supported event types.
Introduction
When you sign up to use Media Recommendations, you work with your Discovery for Media contact to determine the best recommendation models and customizations to use for your site. The models and customizations that you use depend on your business needs and where you plan to display the resulting recommendations.
When you request recommendations from Media Recommendations, the serving config determines which model is used to return your recommendations. You can also filter your results.
Recommendation model types
Discovery for Media offers the following recommendation model types:
Others You May Like
The Others You May Like recommendation predicts the next document that a user is most likely to engage with. The prediction is based on the user's engagement and the current context document.
Default optimization objective: click-through rate
Default serving config: N/A
Available customizations:
- Change the optimization objective to conversion rate
Supported pages for model deployment:
- Detail page. See the
view-itemevent
Recommended for You
The Recommended for You recommendation predicts the next document that a user is most likely to engage with based on the engagement history of that user. This recommendation is typically used on the home page.
Recommended for You can also be useful on category pages. A category page is similar to a
home page, except that you display only items from that category. You can
achieve this using a standard Recommended for You model with filter tags. For example, you
can add customized filter tags (corresponding to each category page) to the
items in your data store. When you send the prediction request, set the user
event object as view-category-page and specify a specific category page's tag
in the filter field. Only recommendation results matching the requested filter
tag are returned. Diversity should be disabled in this use case, because
diversity can conflict with category-based filter tags.
If you want to use the Recommended for You model on the home page, select Home Page
Context as the context event type for the model. Only use the Generic option if you
don't have view-home-page events.
Use the Recommended for You model on pages like Product Detail pages and Add to Cart pages, or even for sending notifications. With the Generic option, you can create a Recommended for You model that can be used on any page.
Default optimization objective: click-through rate
Default context event type: Home page context
Default serving config: N/A
Available customizations:
Change the optimization objective to conversion rate
Change the context event type to Generic
Supported pages for model deployment:
- All. On category pages, you must provide filter tags.
More Like This
The More Like This model recommendation predicts media that is both similar to a context item and that a viewer of the context item is likely to engage with next. The More Like This recommendation is based on the context item and the aggregate viewing history of all users who viewed the context item. The More Like This model is typically used on detail pages or on the home page with a fixed context item.
The More Like This model uses a variety of factors to determine how similar two media
documents are, including the categories field of the media
documents. For best results, media documents that are similar should have
overlapping categories—for example, ["Action", "Comedy"] categories are
somewhat similar to ["Action", "Thriller"], but not similar to ["Drama"].
Default optimization objective: click-through rate
Default serving config: N/A
Available customizations:
Change the optimization objective to conversion rate
Add diversification (supported but not recommended)
Supported pages for model deployment:
Detail page
Home page (requires a context item)
User event requirements:
view-itemormedia-playfor click-through rate objectivemedia-playandmedia-completefor conversion rate objective
Most Popular
The Most Popular recommendation predicts media that had been most popular among all users in recent days. The recommendation is based on the watching or viewing history of all users. You can customize the time window to check the popularity of the documents.
Default optimization objective: click-through rate
Default serving config: N/A
Available customizations:
- Specify the time window in days to check the popularity of documents in the last X days.
Supported pages for model deployment:
- Home page
User event requirements:
view-itemormedia-playfor click-through rate objectivemedia-completefor conversion rate objective
Optimization for business objectives
Machine learning models are created to optimize for a particular business objective, which determines how the model is built. Each model has a default optimization objective, but you can request a different optimization objective to support your business goals by contacting your support representative.
After you have trained a model, you cannot change the optimization objective. You must train a new model to use a different optimization objective.
The Discovery Engine API supports the following optimization objectives.
Click-through rate (CTR)
Optimizing for CTR emphasizes engagement. You should optimize for CTR when you want to maximize the likelihood that the user interacts with the recommendation.
CTR is the default optimization objective for the Others You May Like and Recommended for You recommendation model types.
Conversion rate (CVR)
Optimizing for conversion rate maximizes the likelihood that the user consumes the content up to the conversion threshold defined in the model.
The conversion threshold can be specified in seconds or percentage. For example, if the conversion threshold is set to 25% and the user watches at least 25% of the program, then the conversion objective is met.
Advanced model configuration options
Depending on the model type, there are some other model configuration options you can use to change the behavior of your model.
Tuning preference
Tuning keeps model training optimal as input data changes over time. Set your model to automatically tune every three months, or choose to only tune it manually. The model automatically tunes one time after creation. For more information, see Manage models.
For more information about tuning costs, see Discovery Engine pricing.
Available serving configs and models
Before you can request recommendations from your model, you must create at least one serving config for it. For more information, see Create serving configs.
You can see your models listed on the Models page. Click a model name to go to its detail page, where you can see serving configs associated with that model.