How to build audiences in the Customer Data Platform

Once your Customer Data Platform is up and running and is bursting with data, you are ready to put the data into use. During onboarding training workshops, you have been defining strategies for data activation and how audiences made in the CDP can support your business goals.   

Whether your goal is to attract new customers, keep your existing customers active, transform them from fly-bys to actual buyers, win them back, or a combination of the above, you need to start building audiences in the CDP that in the best way possible express your aim. 

An audience is a collection of profiles within the CDP that are grouped together by one or more behavioral, personal or calculated characteristics.    

In general, we encourage you to play and experiment with the options offered to you in the CDP. This article will walk through how you build audiences, including a typical use-case that you might be able to utilize. 

Build an audience

• Go to the Audience Builder from the menu in Raptor Control Panel and click Build new audience 
• Give your audience a name and a description and click Next

• Now, you have a rich variety of options and filters for building your audiences depending on which data are ingested into your CDP. Below is a description of the general concepts of the Audience Builder. 

• • Choose condition: Did, did not or have:
    - Did and did not conditions are tied to interactional behavior defined in the stream of event based data usually coming in from your website tracking or POS data. Building an audience based on did or did not conditions allows you to use events like buy, visit, add to basket, item click, page view from online tracking and buy in store, or return in store events from your POS system to define your audience. 
    - Have conditions are tied to personal characteristics typically coming from your email marketing system, CRM system or any type of customer database and express data like gender, address/city/country, birthday, bonus status etc. Calculated characteristics coming from your Calculated Attributes or AI Models like the CLV Model typically express values like the sum of all online buys, top visited categories or the predicted lifetime value of the customer. 
    
    

• • Choose source:
    The sources available in the Audience Builder correspond to your schemas from the Data Manager, your Calculated Attributes, your CLV Models and/or your product feed. Sources available will depend on which condition (Did, Did not or Have) you have selected in the previous step. 
  • Set amount
    For Did and Did not conditions you can set a minimum, maximum or exact number of times, a person did or did not perform the action selected under source. For instance "Give me all people who did buy at least 3 times"  
    For Have conditions, this option depends on the data type of the source. 
    Example: For numeric types (like decimals, enum, integer etc.) simple, mathematical statements are available, e.g. Greater than, Equal to, Less Than, Not Equal To, Is empty.
    Note: Greater than and Lower than are available for the data type string. This is due to the fact that on some accounts data like postal code are ingested as strings, meaning the the operators Greater and lower than will in fact work. In most other cases, the two operators do not make sense on strings, so be careful when using them. We always recommend you to make sure that numeric data is ingested as numeric data types (through the Data Manager). 
  • Specify time
    A span of time can be selected to limit the audience, ranging from a handful of hours to the entire extent of available records - that is, 'all time'. By default, this span ranges from present date, but a certain period further back can also be selected using the Between option.
    For Have, this selection is not available - however, if a Calculated Attribute is used, said attribute will have its own time-span built in at creation.
    
    
• Once the basic statement has been constructed, you have a number of options for narrowing or expanding the resulting audience. At the bottom of the block, you can select either And persons who or Or persons who. There is also an + Add filter button attached to every statement.
  • And persons who 
    This option attaches a second statement to the block, acting as a filter. The audience established by this block will be only those who matches both statements. This can be a good place to employ a Did Not statement, for instance.
  • Or persons who
    This option creates a new and separate block - an entirely new statement, independent of the previous one. This new statement can have its own And persons who statements and filters attached, enabling you to precisely define multiple independent groups to combine into a single audience.
  • Add filter 
    This activates a simple filter on the given statement, allowing you to limit the given grouping using customer- or item-specific parameters.  
    
    
• All of these options can be stacked infinitely. It is possible to apply multiple filters to a broad statement in order to create a more narrow definition, group dozens of smaller definitions together into a more useful audience, or both at the same time. If you can imagine a use for it, the Audience Builder can create it.

Example of a Useful Audience

With all these options in mind, here is a specific example of an Audience you may find applicable.

1. Create a new Audience, naming it 'Window Shoppers'. The newly created Audience will default to a Did statement, which suits our needs just fine.
2. For Source, select View Product under Website. Leave the next option as At Least, which is the default, but move the slide-bar to 3. In the time-selector, pick last 30 days. From this, the CDP will gather a list of anyone who've viewed products on your website at least 3 times in the last month.
3. Click the + Add filter button beneath the line to add a simple filter. For the Field, select CategoryPath from the Website-section, and set the second field to Contains. In the final field, you can input a specific category-path for an area you're having a sale in, or a more generic category that can encompass several related areas. For example, for a clothing-store typing in 'shoes' might cover both running-shoes, dress-shoes and the like. Thus, you've narrowed the field to only those who've shown interest within a particular, chosen category.
4. Select the And persons who option from the bar at the bottom. This will add a second line, which will serve as an advanced filter.
5. Change the Did field to Did not. For the Source, select Buy Online from the Website-option. The quantity-slider can be left at the default - at least 1 time. Set the time to the same as the first line - Within last 30 days. This will filter out anyone who've actually shopped at your store.
6. Click on the Save & Refresh button under the Audience Preview to finalize your Audience. The process may take a few minutes, since the list of people who haven't bought anything in the past month is likely to be quite long.
7. Once completed, you will have an audience consisting of people who have looked at, but not bought, specific products from your website. They may be tempted but wavering, or considering options from multiple sources. Either way, a well-placed notification of an ongoing sale or special offer might bring them back with money in hand.

This model is, of course, highly flexible. Want to cast a wider net? Increase the time-frame to three months instead of one, or remove the CategoryPath filter. Want to target a smaller audience more likely to bite? Increase the number of views required in the first line, or specify the Category more narrowly.

The more you familiarize yourself with the many possibilities afforded by the Audience Builder, the more useful applications you are likely to realize. 

A Note on Mixing Sources

Statements and attached filters can draw from different kinds of sources e.g., Website Tracking and Product Feeds. Anything found in a submenu dubbed 'Website', which includes things such as Buy Online, Product View, Subtotal or BrandID, is part of the Website Tracking. This data is based on accumulated performance, with all past incidents being part of the calculations. This means that, for example, if the ItemID for an item is changed, an audience set to include those who shopped for the original ItemID will encompass everyone who bought that product before the change was made and only them.

Product Feeds, meanwhile, only draw from the present state of your database, as of the most recent sync. This includes information such as Current Stock, Color, SalePrice, Name, etc. For these calculations, changes to the data are irrelevant - only the most recent data is used, with every new sync overwriting the old records. This type of data is mostly utilized by filters, and will thus often be mixing them with the accumulated data stemming from Website Tracking.

This can create some issues if you attempt to build certain, specific types of Audiences. For example, building an Audience consisting of everyone who bought a product at below a given price - which is data drawn from the Product Feed - will search back through all previous buy-records, but use the current pricing only. The purchases it finds and builds the audience from may have been made at a different price altogether, if the product has dropped or risen in base price since. Likewise, filtering based on a previously-popular product that has since been removed from the shop listing due to being discontinued will not work, assuming you search for a specific identifier such as Name of ProductURL - if the product isn't part of the current Product Feed, the system will not be able to identify occurrences of it in earlier transactions. You could still search for it based on something broader like Brand, however, assuming said brand remains part of the current Product Feed.

Consider how to work around this when drawing from different sources when you create your Audience.