Fallback techniques can enhance your Einstein recipe-building strategy and ensure that
a recipe returns enough items to fill the recommendation slots designated by a
campaign.
This table lists possible fallback options to use, their benefits, and things to consider
when using them in a recipe.
| Fallback |
Benefits |
Considerations |
| Any Eligible Item ingredient |
- Acts as a fallback for other ingredients.
- If there isn’t enough user data to provide a full list of
recommendations, this ingredient includes items that meet the minimum
data requirements for the target recipe.
- Populates only enough recommendations to meet the minimum number of
promoted items for the campaign.
- Compatible with all other ingredients.
- Eligible items obey existing exclusions, inclusions, and
variations.
|
- For a product to be recommended, it must meet the minimum requirements
(name, ID, URL, image URL, price, and an inventory of null or not
zero).
- If the recipe doesn’t include exclusions, inclusions, or variations,
this ingredient returns a random set of items.
|
| Collaborative Filtering ingredient |
- Has trending built in.
- Defaults to showing trending results when there’s no user or item
activity during the configured lookback period.
- Trending results are identical to using a Trending ingredient with a
weight value of 1.
- Eligible items obey existing exclusions, inclusions, and
variations.
|
- Built-in trending fallback applies only to the collaborative filtering
ingredient.
- You can’t customize trending parameters. Trending is hard-coded to
recommend products by most purchased and recommend other catalog objects
by most viewed.
- The trending in Collaborative Filtering ingredients stacks with Trending
ingredients in recipes. This stacking is equivalent to having multiple
trending ingredients in the recipe, each with potentially different
weights. In recipes containing three or more ingredients, and no user or
item activity has occurred within a specified lookback period, stacked
Trending ingredients can overpower other ingredients in the recipe.
|
| Trending ingredient |
- Enables setting a trending weight (1–4) using a graduated slide.
- Eligible items obey existing exclusions, inclusions, and
variations.
- Recommends products by most revenue, most purchased, most time spent
viewing, or most viewed.
|
- Provides trends using statistics from only the past 24 hours.
- No configurable lookback period.
|
| Enable trending product fallback option |
- Checkbox works for the entire Einstein recipe.
- Populates only enough recommendations to meet the minimum number of
promoted items for the campaign.
- Enables you to evaluate products for trending over a specified number of
past days. The maximum setting for evaluation is for products from the
past 30 days.
|
- Shows trends only by purchase count for products and view count for
other catalog objects.
- Ignores exclusions, inclusions, and variations.
|
Precautions for Email Campaigns
When using a recipe in an email
campaign, the system checks whether the email contains blank recommendation slots. By
default, all Open Time email requests perform a Trending lookup (by views over the past
week) to fill empty slots with items. If you don’t want to show recommendations in an
Open Time email, you can override the default behavior. On the item block, deselect
Enable trending product fallback for the recipe.
The Trending lookup used by an
Open Time email request ignores inclusions, exclusions, and variations.
