Community
When would you not âEnable API Accessâ?
When using the Lytics audience builder, there is a checkbox to Enable API Access. This option is not enabled by default for security purposes, but you may need to enable it for personalization use cases involving web, Facebook, Google Analytics, and more. So you may wonder, "why not enable API access for all of my audiences?"
A reason to keep API access disabled for some audiences is if you don't want to automatically sync them with your integrations. This will help to significantly reduce clutter in your downstream tools. Audiences with API access enabled are stored in a cookie, which Lytics uses to enable personalization campaigns. You can ensure that the size of the cookie is optimized and only relevant audiences are stored by keeping API access disabled for audiences that you do not want to target.
Posted by Mark Hayden over 1 year ago
How do you A/B Test within Lytics Audiences?
A/B testing within Lytics Audiences can be used to understand how Lytics Audiences are performing. Using a random generator generator, Lytics assigns 2 random numbers to each user profile calls _splits_. These fields are named `Random Split 1` and `Random Split 2`.
To use these fields, log in to Lytics and navigate to the **Audience** -> **Create New Audience**, then search for **Random Split 1** or **Random Split 2** within **Custom Rule**. These fields evenly assign a value from 0-99. To create an audience of 1% of the population, select a number (47 in the example below) and use the `equals` operator.
![A/B Split1 ](//images.ctfassets.net/p3327y1wyjsx/4TPubx4bTdhr9o1NiwxDHo/ca89e5f9234476fd7e4e0fdd6fe8f901/Screen_Shot_2022-02-07_at_3.06.49_PM.png)
To create random groups smaller than 1% of the population, you can leverage both of the split fields. In the example below, we create an audience of roughly 0.5% of the total user base by defining `Split 1 = 47` and `Split 2 < 50`.
![A/B Split2](//images.ctfassets.net/p3327y1wyjsx/1hdcMdaekGBbPaJqyFhm7L/a93dd6ebd7c805216f388b16451a70bc/Screen_Shot_2022-02-07_at_3.07.45_PM.png)
\*Note that this field is approximate and won't necessarily be perfectly distributed.
Posted by Mark Hayden over 1 year ago
How long does it take for my audiences to update in Lytics?
Audiences membership update in real-time as data is received by the Lytics platform. As event data is received from your multiple channels, Lytics determines if this new data effects audience membership, the Lytics platform will move the user into and out of the audiences that you have defined.
Posted by Mark Hayden over 1 year ago
What are characteristics?
Characteristics are a type of pre-defined audience, meant to represent factual information about a user with the sole purpose of enhancing audience segmentation capabilities. For instance, the âHas Emailâ characteristic is comprised of all profiles that have some value for email. This can be viewed as an aggregate, how many users are a member of âHas Email,â or on the individual profile level.
It is best to think of characteristics as a high-level analytics metric as well as a set of predefined building blocks for your custom segments.
Posted by Mark Hayden over 1 year ago
Why my audience is not being returned in the Lytics Javascript tag?
You will need to check the Enable API option in the audience editor of the audience you would like to see returned to the Javascript tag.
Posted by Mark Hayden over 1 year ago
Why are my audiences different sizes on different pages?
Lytics regularly caches the sizes of audiences to increase the responsiveness of the interface. If you have recently edited a segment, or have just sent large amount of event data, the dashboard and summary pages may have an out-of-date audience size. To see the most accurate audience size, view the audience and click the pencil icon to edit. The audience editor will always show the most up-to-date segment size.
Posted by Mark Hayden over 1 year ago
What are pre-defined audiences?
All Lytics accounts comes out-of-the-box with some universally useful pre-defined audiences.
- All users
- New users
- Highly engaged users
- Currently engaged users
- Previously engaged users
- Disengaged users
- Unscored users
- These audiences are immediately accessible in the audience section of the Lytics interface.
Lytics also offers pre-defined audiences called "characteristics" which you can see on the dashboard as metrics. These include users who are highly engaged in the most active topics in the account's content taxonomy. And some channel specific behavioral audiences.
Like any other audience, pre-defined audiences can be used as building blocks to incorporate into other segments that you may build.
Posted by Mark Hayden over 1 year ago
How can I track user lifecycle (e.g. highly engaged, at-risk)?
Lytics tracks user behavior trends on the reporting dashboard. The section called "Total Audience Characteristics" contains a section called "Behavior", which includes engagement trends such as "High Usage", "Binge User", "Likely to Re-Engage", "At-Risk" and more. You can filter by these behaviors when creating audience segments.
Posted by Mark Hayden over 1 year ago
What does "self-learning" mean?
Lytics Behavioral Scores are meant to adapt over time to the behavior of the users in the account. We do this by updating the benchmarks (this includes the statistical models and parameters) used to calculate the Behavioral Scores every three days. In doing so, we can ensure that the Behavioral Scores stay âfresh.â
Posted by Mark Hayden over 1 year ago
What is machine learning and how does Lytics use it?
Machine learning is a form of artificial intelligence that involves applying techniques such as pattern matching and computational statistics to large amounts of data to predict events. Simply put, itâs training âmachines to learnâ -- hence the name. Lytics applies machine learning to your customer data to help make predictions around what people may do next: make a purchase, leave the brand, have affinity for a certain type of content, go on a binge buying spree, and so on. This is how we uncover highly valuable audiences such as âLikely to Buyâ that you can narrowly target in online advertising, email marketing and web personalization campaigns.
What makes our machine learning unique from that of other customer data platforms is we look at more than 125 factors about a personâs behavior to predict what theyâll do next. Other companies look at only a handful of aspects of peopleâs behavior. Technically speaking, Lytics implements an ensemble of statistical techniques to describe a person across a set of linearly independent behavioral dimensions. Techniques include random forests, logistically weighted moving averages and quantile estimation.
Posted by Mark Hayden over 1 year ago