In an effort to give users a better starting point, a few sample audience building ideas and details are contained within this article. The ideas for these audiences have been sourced from a number of use cases we've handled over the years and we are able to achieve within the confines of a sample dataset. Your use cases and data may be completely different, the goal here is to get the creative juices flowing.
As a reminder, all Audience Builder apps use Attribute Builder apps as the source. If you are unsure of what an Attribute Builder is please refer to the Attribute Builder documentation or ask your organization's Audiences Solution owner if you do not know what Data Layer to use.
Resultant datasets are a list of visitor IDs in a back-end table with a date column available to be synced with downstream systems, e.g. DSPs.
The sample audiences contained within this article are being built off of the Sample Attributes detailed in a separate article.
Common e-commerce website audiences that marketers need to have available center around product-related information, such as brands. This section walks through how an audience may be built to determine the visitors that interacted with a brand but did not purchase within the last seven days.
It is common for certain marketing activities, e.g. retargeting, related activities for a user to want to know what visitors interacted with the website in a certain way and haven't purchased recently.
In the sample below there are two sets of attributes being used to create an audience:
- E Comm Last 7 Days
- Brands 1 Day
These input nodes are connected directly to the Audience Builder node and made available for defining the conditions of your audience.
Within the conditions a logical AND is being used to combine the attributes, essentially saying:
"Give me all the IDs that interacted with Sony brand yesterday and have not purchased in the past seven days."
Retention - 90 Days, retention is a metadata value passed to downstream systems telling them how long the audience should be persisted.
Machine Learning Model Audiences
After a machine learning model is built the results can be in various forms, such as a raw numeric score or labeled categories, depending on the type and purpose of the model. Attributes derived from these are either passing the data point as-is or providing some categorization and labeling of the raw scores. The audiences built from ML scores typically are bucketing the results, i.e. High, Medium, Low, and possibly adding some behavioral related attributes to further refine your audience segments, e.g. visitors that have not purchased in the last 30 days and have been labeled as a high propensity to purchase.
In the sample below, a purchase propensity machine learning model was built where the data scientist provided the users with a table of raw scores ranging from 0 to 1 and attributes were built to set thresholds to define 'high propensity to purchase' and 'low propensity to purchase'. These attributes are being to derive audiences with retention set at 14 days.
Most e-commerce websites will keep cart and purchase-related audiences in their commonly available stock. Audiences like cart abandoners and repeat purchasers are typically desired.
In this example, we've built several e-commerce-related attributes and we want to build an audience that had any type of e-commerce related action yesterday. The keyword ANY in the right value allows the app to locate any IDs where there is an 'E Comm Last 7 Days' attribute.