Online businesses aim to meet customer expectations while competing in the market. Advanced technologies help them achieve these goals through cloud services, customer data platforms (CDPs), and machine learning (ML). This also enables the analysis of a considerable amount of raw data, turning it into meaningful insights that help businesses serve their customers better in real time.
In this sense, online businesses either build in-house solutions or use third-party solutions to understand their customers and create a better experience that helps build and maintain loyalty. Retaining customers requires understanding their needs, behaviors, and motivations with various tools.
This guide aims to answer the following questions:
What kind of algorithms are available in Predictive Segments?
How can you use propensity scores that you calculate outside Insider One?
How can you understand customer behavior?
Customer data platforms (CDP) collect events, user, and product attributes on a website or application and store them in their databases. This data is used to identify new or returning users and their interactions with categories, products, and services. Segmenting these users is a great way to understand the intent and behavior of a group of people, tracking the paths that lead to purchases, cart abandonment, churn, or loyalty.
Although segmenting a group of users is helpful for marketing promotions and showing recommended content and products, the identity factor might be missing. Segments are high-level targeting, and personas are low-level targeting, like a micro-segment that drives better results in marketing campaigns.
Additionally, data is required for the algorithms to work. The more data there is, the better the algorithms will make predictions. Therefore, if the data is not sent after the customer starts using a predictive segment, the algorithms will not work.
How does Insider One use data in algorithms?
Insider One uses online and offline data in predictive algorithms.
Likelihood to Purchase can only be calculated with real-time data.
User Engagement can be calculated with online data to assess the customer's overall online engagement with your platform.
Customer Lifecycle Status (CLS) works directly with both online and offline data for purchase event data.
Both online and offline data can be used for Discount Affinity algorithms. While product page view event data should be collected via online and offline channels, purchase event data can flow from online or offline channels.
Attribute Affinity can be calculated with the customer's online product page view event.
How does Insider One help with the Predictive Segments?
Insider One analyzes user data and historical trends to define users' interests and affinities and predicts future intents using machine learning methods. For example, if a user interacts with sweatshirts during a session and views the red ones most often, they are likely interested in red. The vendor could display the red-colored clothing items at the top of the catalog on the next visit.
Machine learning algorithms can be trained to understand the differences among shirts, shoes, and jackets, as well as their attributes, such as colors and brands. In addition to category visits, engagement levels with messages or emails, discounts and promotions, visit densities and durations, cart adding, purchase frequencies, and other data points can also help identify micro-segments.
What kind of algorithms are available in Predictive Segments?
Insider One offers the following algorithms:
Reach out to the Insider One team to enable the respective algorithm(s) that you would like to use.
How can you use propensity scores that you calculate outside Insider One?
Insider One's Predictive Segments are calculated from the events and attributes you send. If your team already produces its own propensity or likelihood scores elsewhere, for example, a churn-risk score, a customer lifetime value band, or a next-purchase propensity produced in your own data warehouse or data-science environment, you can bring those scores into Insider One as custom user attributes.
Once imported, a score behaves like any other user attribute:
You can target it in the segmentation builder and combine it with Insider One's own Predictive Segments and standard conditions using AND/OR logic.
You can use it as a journey condition in Architect, for example, entering users into a journey with the On Attribute Change trigger when their imported score crosses a threshold.
Imported scores refresh according to the schedule of the integration that sends them, so treat them as batch-updated attributes rather than as live, per-event predictions. Insider One stores and segments on the exact value you send; it does not recalculate or transform your score.