Segment insights
Segments aren't used by for items like measures, but they can also be exported and used in other settings. For example, a segment can be exported to Customer Insights - Journeys, where it can be used in customer journeys. As you use segments more, you might want to understand how the different segments relate to each other. This might include understanding how they're different and what they have in common. This can help make better decisions related to which segments you should be targeting across different promotions, campaigns, customer activities, and so on. For example, as you plan your marketing strategy for the next six months, you might want to identify how much customer overlap there is between the segments that is targeted in the campaigns. By identifying this information, organizations can more effectively plan how the work with customers.
To help with this, Customer Insights - Data includes multiple features for examining and expanding segments. feature called Segments Insights.
The three main tools available for insights and expansion are:
Customer Overlap: Lets you visualize shared members between segments
Segment Differentiators: Lets you see what distinguishes segments from each other or from the rest of your customers.
Find Similar Customers: Lets you create a segment based on an existing segment that includes customers with a profile similar to the customers in the selected segment.
Finding Similar Customers can be initiated from an existing segment. The overlap and differentiator insights are part of Segment Insights that can be accessed from the Insights (Preview) tab.
Segment overlap
With segment overlap analysis, you can see how many customers are common to two or more segments. For example, a segment of frequently contacted customers might overlap with a segment that contains customers that are satisfied with your service or product. Within the analysis, you can further break down the attribute level filters to break down the numbers at a field level.
An overlap analysis can contain between 2 to 3 segments to help visualize which customers are shared between them. Additionally, up to five fields of interest can be used to better analyze overlaps at a field level. Once created, analysis can be modified as needed by adjusting the fields selected to see how the changes affect the results. After each change to the analysis criteria, the data automatically refreshed so you can view the results.
Each analysis contains a visualization report that shows the number of overlapping customers you have based on the segments included in the analysis. A table at the bottom identifies the total members for each list and the percentage of the total results. In the image below, you can see that we have 527 overlapping customers between the two segments included in the analysis that represents only 18.9% of the total combined customers from both segments.
Any fields that you included during configuration are represented with corresponding tabs at the top of the report. Each field contains a filter drop-down to let you filter the data down into more targeted areas.
Segment differentiators
Segment differentiators help identify what differentiates a specific segment of customers from the rest of your customers or from another segment. You identify the segments that you want to compare along with any fields that should be used as the differentiator. It's also possible to identify if any existing measures should be included as well. After you save the analysis, it will refresh itself and once the refresh is complete, you'll be able to use it.
A differentiator analysis includes two tabs.
Attributes tab: Lists profile attributes considered as differentiators.
Measures tab: Lists differentiators.
Each tab displays a ranked list of all the differentiators, sorted by difference score. The difference score represents the degree of difference of an attribute between two segments. The higher the difference score, the more the attributes differ between the two segments. When you select a score, it opens the difference score pane that provides distributions of values for that attribute.
Find similar customers
Customer Insights - Data find similar customers feature uses artificial intelligence to identify similar customers across your customer base or in other segments. This can be useful to help expand successful customer segmentation to other areas. For example, your company might have just run a successful product launch in a specific region and is looking to replicate that success in another region. Finding similar customers lets you identify customers that likely have the same customer profile in the next region you want to target.
To identify similar customers, you must have at least one segment created. You then can select the segment that you want to work with and choose Find similar customers. A new segment is created based on the profile defined in the selected customer segment. You can review existing or add any new fields that you want to use to define your new segment. They define the basis on how the system finds similar customers to the source segment you selected. The system tries to identify some recommended fields by default and automatically exclude any fields that can reduce the model performance.
These excluded fields typically include:
Fields with the following data types: StringType, BooleanType, CharType, LongType, IntType, DoubleType, FloatType, ShortType
Fields with a cardinality (the number of elements in a field) of less than 2 or more than 30
The key part of identifying similar customers is to specify the customers to include in the new segment. You can include all customers expect those from the source segment, or only customers from a specific existing segment. By default, customers from the source segment aren't included. You can include those customers by selecting the Include members from source segment in addition to the customers with similar attributes check box.
When you create the new segment, define how many similar customers to include by specifying a percentage of customers to include. By default, only 20% of the target audience is selected. As the precision percentage is adjusted, the segments members are either expanded or reduced. How similar the customers in the new segment are to the original is also affected. For example, reducing the number of records to include decreases how many customers are included, but the customers are similar to the source customer base. Increasing the number includes more customers, but they might not be as similar to the original customer base. After you modify your settings, select Run at the bottom of the page to analyze the dataset.
Binary classification is used to score the customers that are returned in the similar segment. The score is based on the similarity to customers in the source segment.
Below is a breakdown of how the scoring is defined:
Below 0.55: Customers are not similar to customers in the source segment
Between 0.55-0.7: Customers are similar
Between 0.7-0.85: Customers are similar
Between 0.85-1: Customers are similar
After the new segment is created, use it as you would with any other segment. For example, export the segment to use it in another application or building a measure from it.