Ever wonder what the data provider network looks like?

Audience buying has revolutionized digital marketing. Media buyers have the ability to construct audiences using hundreds of thousands of data consumer attributes -- what they buy, where they shop, how old they are, what their interests are, their financial information, and more. Powering the audience market are data brokers. These brokers collect, curate, and codify billions of interactions happening online and offline and translate them into buyable audience attributes.

When the data is captured it has very little value or practical application. Data brokers create value by organizing interactions into attributes that marketers can buy (example: auto-intender). Even more, the data's true potential is realized when that data is made available in an audience buying platform -- or more commonly known data a Data Management Platform (DMP). A DMP has been come essential software for audience buyers allowing them to create, analyze and distribute audiences with a few button clicks.

In the world of audience buying, the DMP is king. To truly understand where the DMP gets its power, we have to understand the relationship between the data broker and the DMP. What broker data is in a DMP? Are there exclusive DMP-broker relationships? Which brokers are distributed to which DMP?

Conceptually the relationship between broker and platform is simple -- data broker A calls a DMP and tells them about the data they have. The DMP agrees that the data would be valuable to its users and integrates data broker into its platform. Admittedly, this is a gross oversimplification. In reality, this simple exchange has occurred thousands of times. The result is a complex network of relationships that make it nearly impossible to understand whose data live where -- unless there was a visualization to help us.

Interpreting the map

In a network graph, there are nodes (the dots) and edges (the lines). In the case of this visualization, the dots are brokers and DMPs and the lines represent a relationship between them.

The Broker-DMP network show hidden relationships and reveal critical nodes in the audience economy.

There is a ton that we can learn from the network. We can generate a conceptual map of what relationships exist. We can learn which broker relationships are exclusive or shared. For the purpose of this post, I wanted to summarize some of my initial observations.

Top line understanding

• The map shows connections to and from 10 major Data Management Platforms to 709 unique data providers
• Average DMP has 129 data brokers in its platform
• Neustar, V12 (10), and eXelate (9) have the highest connectivity across all DMPs
• There are 1,289 total connections between data provider and DMP

Consider the complexity

The above visual just shows ten DMPs. The data was sourced from websites that may or may not be updated at the time of this post. Consider that there are hundreds of DMPs and thousands more data providers not pictured here.

Even more mind-bending, consider that each data broker creates an average of 760 data attributes. If we could zoom in on this network, we would see an infinitely more complex network of hundreds of thousands of buyable audience attributes. Below is a visualization of just one DMP's attributes.

The scale can be intimidating -- but scale does not equal quality. Being able to create and buy audiences using +100K attributes, does not mean that the audience you create is the actual group of people who will see your ads.

In future posts, we can dive into attribute usage, frequency, and accuracy. For now, just know that there is an rich, complex, and interconnected marketplace of data that allows programmatic buyers create audiences.

Notes on the visualization

Starting with the DMPs: Using the most commonly found DMPs across 'Top 10' lists I identified the major platform providers: Adobe, DataXu, Google Audience Center, LiveRamp, Lotame, MediaMath, Nielsen, Oracle (BlueKai), SalesForce (Krux), and The Trade Desk.

Gathering the brokers: Somewhere within the company website for each DMP is a list of data providers. Using the information on this page, I generated a list of data brokers for each DMP.

Connecting the dots: The visualization was built using igraph, network, ggraph packages in R.