Click to learn more about author Jonathan Lee.
In the past year and a half DATAVERSITY readers have been exposed to new Customer Data Platform (CDP) offerings from vendors like BlueConic, RedPoint Global and others. But what is a CDP? First introduced around 2013, CDP is a concept defined by the Customer Data Platform Institute as:
“Packaged software that creates a persistent, unified customer database that is accessible to other systems…the CDP creates a comprehensive view of each customer by capturing data from multiple systems, linking information related to the same customer, and storing the information to track behavior over time.”
Some readers may object to the introduction of yet another marketing acronym. After all, one may ask, aren’t the capabilities of a CDP covered under the capabilities of preexisting terms like DMP (Data Management Platform), Data Lake or Data Warehouse? Not quite.
Differences between a CDP and other Systems
According to CDP vendors NGDATA and Fospha, CDPs primarily differ from DMPs in four key respects: function, operating data types, user profiles and data capture. To illustrate, CMPs gather first party data (such as what’s stored in your CRM) while DMPs primarily work off of third party data (information typically collected and sold by an aggregator). DMPs are therefore generally used to reach target customers within an advertising campaign while CMPs are more often leveraged as a means to interact with existing customers and prospects (such as to extend special offers). CMPs are also able to collect and analyze richer, individual data, such as social and offline data while DMPs use data collected from anonymized cookies and so lack visibility into individual characteristics. Data stored in DMPs is also rarely held for longer than 90 days, whereas CMPs analyze data, such as customer retention information, over much longer periods of time.
A CDP is even further removed from a Data Lake or Data Warehouse. Both of the latter terms refer to enterprise-wide initiatives that are typically far removed from the needs of business practitioners like marketers. Data Lakes are simply large repositories of data collected from various source systems. They play an important role by serving as a single repository for enterprise data but that data remains in the format produced by different “tributary” sources, making it difficult to use for analysis or derive insights. A Data Warehouse takes this a step further by providing a means to harmonize and standardize data for analysis but is not intended for customer-facing use as information stored within is generally only updated in periodic batches, rather than on a real-time basis like in a CDP.
Benefits of a CDP
Without getting into the debate over what a single customer view is, or whether it even exists, it’s not hard to understand why a CDP would be valuable to any department touching customer data, particularly marketing.
As touched on above, the traditional method of Data Integration through ETL involves IT resources connecting to data sources, extracting data and performing data cleansing, harmonization, customer ID matching, etc. before loading the data into a warehouse. This is a huge effort requiring time, costs and specialized data analyst skills. And most of the time, the results are not especially timely, particularly given the growth in volume and velocity of enterprise data.
At the risk of oversimplification, CDPs get around this process by serving as a “data virtualization layer.” Just as traditional virtualization creates a virtual (rather than physical) version of something – hardware, storage, etc. – data virtualization means that the CDP acts as a unified data layer that is abstracted, without being constrained by the data format or schemas of the data source. The CDP delivers only the data a user has chosen, without mass data duplication. This doesn’t require a data integrator to setup a predefined schema or fixed field mappings for different event types. The data definition is flexible and can be updated at any time.
The precise customer segmentation that emerges from the deployment of a CDP means marketers can deliver a granular, targeted customer experience without having to force customers into predefined personas. Just as importantly, a CDP allows marketers to run analyses on customer data with no technical knowledge and without having to rely on IT for support. At a minimum, this means running far fewer SQL queries. Finally, a CDP works towards providing the mythic single view of the customer by connecting various customer interactions, such as if a customer visited a physical storefront before making a purchase online.
Criticism of CDPs
No discussion of CDPs would be complete without at least a brief mention of the criticism leveled at the term in past few months by vendors like SAS and Resonate. Much of this criticism has less to do with the stated goals or underlying technologies of CDPs and more about the hype that surrounds them. In particular, there is the, undoubtedly true, argument that some CDP vendors are not necessarily bringing anything new to market, instead simply rebranding existing solutions with a “sexy” new tagline. There is also criticism of the notion that deploying a CDP will truly obviate the need for IT involvement. Commentators have pointed out that a CDP is unlikely to be a “set it and forget it” solution, IT resources will likely still be required, particularly to help with integration into an organization’s existing technology stack
Perhaps more importantly, the point made by Resonate COO Andy Hunn is that even if CDPs do what they claim to do, identify customers, this information is of only limited value. Far more important is to understand why an individual or organization has chosen to become a customer, or not. To illustrate, it’s great that Sally has bought a burger from our burger stand but it’s far more useful to know why she did so – convenience, taste, something else? – and why Jose has not, because that’s the kind of information that will allow us to double-down on our value proposition, improve shortcomings and ultimately increase sales.
Hype aside, a CDP does offer clear benefits if executed properly. However, such a system doesn’t necessarily have to be purchased from a vendor, there is nothing stopping you from building out one internally. A decision just needs to be made about whether the development and maintenance time is worth the cost savings of going with a homegrown solution. Significant development time will need to be devoted not just to the core CDP functionality but also to building connectors for each data source.