The role of forwarding and modeling Behavioral Data within the Modern Data Stack

Share

Data teams are measured on their ability to empower users across the business to make data-driven decisions. Whether that is providing them with knowledge and insight, or equipping users with the right data in the tools they already use to discover and action their own findings.  

Every team has different tools and use cases for behavioural data, from marketing teams wanting to power their marketing automation to  product teams wanting to understand and analyse customer journeys. With such varied use cases across a business, a data team must consider the best method to send data to these downstream destinations. 

With the cloud data warehouse sitting at the heart of the Modern Data Stack there are fundamentally two ways to empower these platforms and users with behavioral data. The first is forwarding events directly to downstream tools via a server-side tagging solution such as Google Tag Manager Server Side (GTM SS). The second approach is to publish modelled data from the data warehouse via a Reverse ETL such as Hightouch or Census. Both forwarding and publishing data both have their associated advantages, lending themselves to executing different use cases. 

With these two methods in mind there are two questions central to choosing the right way to power the use cases across your business. 

  1. Would the use case benefit from processed behavioural data from the data warehouse? 
  1. How important is the latency of data from the event occurring to action? 

Prior to the rise of Reverse ETLs, Customer Data Platforms (CDPs) had made it increasingly commonplace for businesses to default to forwarding behavioural data across all of their platforms, often treating the data warehouse as another destination, as opposed to creating a single rich data source for downstream tools.

Publishing modelled behavioral data allows brands to create richer data sources for downstream tools. Behavioral data has been modeled to filter, aggregate and combine it with other datasets within the warehouse before data is published to other tools. 

Let’s take the example of a data team wanting to empower their marketing team to create highly-targeted newsletter campaigns in a customer engagement platform such as Iterable. We could build a deep understanding of users in the warehouse by identifying behavioral signals for segmentation such as content engagement that drive campaign engagement. At first this could be simple if-else rules, while overtime building algorithms to find hidden patterns in behavior. We could then publish these highly targeted segments to tools such as Iterable or Braze via our Reverse ETL. 

From this example we can see that publishing modeled data provides end users with the most complete understanding of the customer, allowing them to make the most informed decision when actioning the data and creating segments. The end result – greater levels of engagement, and therefore revenue, from campaigns.

When we examine the use cases for forwarding data via server-side solutions, they should often be reversed for individual events where no data processing is required. If we again take a look at the application within an engagement platform, forwarding behavioral data is best reserved for powering triggered behavioral campaigns or in-session nudges to influence the customer’s journey. The key objective for a marketer is to convert the customer with automated campaigns. There is no requirement for published data to power this use case and the timeline of delivery is critical to convert the customer. 

Though published data is where teams unlock the most value from the data, forwarding data can be a good starting point to help data teams allow users to self-serve in the tools they already use, until the case can be progressed. If we take the example of a marketing team wanting to optimize advertising spend, the first step could be to simply forward the data to their chosen advertising platform for conversion tracking alongside creating a simple attribution model to measure performance by channel. 

The long-term objective would be to build advanced audiences in data warehouses alongside creating advanced custom multi-touch attribution models to truly measure the impact of ad spend.

Shifting the mindset to consider the end use case of the data is an essential step to empowering both users and downstream tools. It’s essential to remember whether you choose to use forward or publish data to power your use cases that its effectiveness should be measured by analysing the data in the warehouse. 

What are the use cases across your business and are you currently powering them with the right method? 
Be sure to check out our release article on Snowplow’s support for Google Tag Management server-side for forwarding event data or see how you can put data into action with Reverse ETL.