CDPs are only scratching the surface of their potential
Instead of solving core challenges, the majority of companies are focused on the application of data rather than serving as an integrated data platform according to Michael Katz, the CEO and Co-founder of mParticle.
A few years ago, nobody had heard of customer data platforms (CDPs). Now we are in peak hype cycle and it seems lots of vendors want to ride the wave.
The perfect storm of device fragmentation, the explosion of marketing tech tools, plus the shift to identity, and the rise of the API economy are a few of the factors that have led us to where we are today. A single customer view to drive cross-channel personalization is something almost every brand is currently chasing. As Forrester analyst Joe Stanhope writes in a new report (Forrester subscription required), “Brands need a modern data fabric to support high stakes customer engagement.”
The problem that brands are facing today, according to Stanhope, is that the CDP category is now a mile wide, and many vendors, only an inch deep. Rather than solving core data challenges, the majority of companies currently billing themselves as “customer data platforms” are, instead, focused on the application of data, rather than the enablement of other vendors. They are more point solution than they are a platform.
Therein, lies the confusion. If every company that ingests and stores customer-level data, and offers a unified customer profile, is a CDP, then almost every martech vendor can (and probably will) call themselves a CDP. As Stanhope writes, “The lack of structure and go-to-market rigor in the CDP market today makes it difficult for marketers to understand potential benefits, identify prospective vendors, and make the business case to invest.”
Some historical context
While specialization and disaggregation are not new trends in maturing technology markets, this isn’t the first time we have seen vendors in a different part of the stack attempt to attach themselves to certain buzzworthy data offering.
A very similar phenomenon happened in the advertising technology space as the data management platform (DMP) ecosystem started to take shape back in 2010. The DMPs came to prominence as a solution for marketers and publishers to regain data leverage over the ecosystem of activation-centric intermediaries (ie ad networks, and demand-side platforms (DSPs)). As the DMPs became the talk of the industry, every major DSP at the time soon was offering a bolt on DMP offering. The rationale they gave was that it was a key feature to enable effective paid media activation, rather than a product onto itself. The problem with this rationale is that it ignores a strong historical precedent, plus the presence of a stand-alone data platform separate from marketing and measurement vendors have been around for a very long time and is required to mitigate channel and partner conflicts.
A better view of the market
Irrespective of the current market confusion and fragmentation, there are essentially only two kinds of CDPs: 1) ones that focus on creating a foundational, customer-centric data pipeline, and 2) ones that focus on the front-end, application layer.
Stanhope refers to foundational CDPs as “data-pipes-oriented CDPs”, we think that the term “foundational” is more appropriate, as the emphasis is on helping brands implement a proper data and identity strategy, and easily integrating that data into a constellation of partners, serving as the anchor system. This contrasts to the many flavors of “application layer” CDPs (which Stanhope in his report refers to as “measurement-oriented,” “automation-oriented” and “orchestration-oriented” vendors) whose emphasis is on reporting and activation capabilities.
Foundational CDPs like mParticle, Segment or Treasure Data are not “yet another tool” in the stack, they serve as the anchor system in an industry where there hasn’t historically been an anchor system. The focus is on addressing core challenges associated with data collection, control, and hygiene; organizing that data through sophisticated identity resolution capabilities; and orchestration of data flows, improving the data that’s available to all the other tools in your stack. Simply put, they help you get more value out of your vendors, they do not replace them.
The buyers of these solutions typically serve in a product-oriented capacity with a focus on marketing technology and customer experience. They may be highly technical and their purview into the stack extends well beyond any one use case. This contrasts significantly with buyers of an application-layer CDP who are usually less technical, or cannot get engineering support and are focused on a singular use case.
Even if, as Stanhope suggests, the CDP Institute’s current definition of a CDP is too broad to be useful, we shouldn’t throw out the baby with the bathwater. Instead, companies should be clear with what they are looking for. If what someone is looking for is modern customer data infrastructure (and not a particular marketing or analytics application), below are five important ways to separate real CDPs from those pretending.
What constitutes a foundational CDP?
At a minimum, a foundational CDP should meet the following five criteria:
1. Flexible, Real-time data ingestion and export tools
Why should you care: Every organization has unique data infrastructure requirements to support their customer experience; a foundational CDP should offer an array of APIs, SDKs, and other tools that allow you to get data to & from in real-time to keep pace with the speed of the customer. Additionally, it should be able to standardize data across these systems while having the controls to solve for nuanced differences in data schemas and taxonomies.
What to beware of: Look out for CDPs with limited data collection capabilities or limited outbound integration support. Anyone can transfer data by brute force, and not all integrations are created equal. Point to point integrations are a fine band-aid, but they don’t get to the root of the issue. One key piece of advice would be to make sure the system is API-driven rather than flat file or CSV export.
2. Ability to standardize data at ingestion for downstream activation partners
Why should you care: Data hygiene is a problem for even the most technically sophisticated organizations. The ability to standardize tagging across ingestion points or customize naming and formats for a specific downstream system is a core requirement for enterprise data usability. And the idea time to standardize data is at the point of ingesting before it gets into other systems or tools.
What to beware: One-off manual transformations or bolted-on MDMs.
3. Deterministic, customizable identity resolution
Why should you care: At the core of Integration is Identity. To address the entire customer journey, from first touch through transaction, across multiple devices, partners, and channels; a proper solution should take a non-prescriptive approach to identity resolution. Different systems may require different identity information (anonymous vs. known, email vs. CRM identifier); a CDP needs to be able to intelligently unify data regardless of origin or intended destination.
What to beware of: A one size fits all approach to identity resolution, or even worse, no identity resolution capabilities at all. Whether it’s a rigid framework that doesn’t account for various forms of user identity or systems that can only address known users but not anonymous traffic, these solutions will lead to user record mismatching and poor data hygiene which will have a cascading downstream impact.
4. Real-time and bi-directional integrations
Why should you care: A foundational CDP should be built on real-time streaming architecture, full stop. They also need to get data to vendors as well as offer bi-directional data flows that give you feedback loops from various downstream systems. That way, for example, a marketer can intelligently optimize conversion paths. For example, if a user converts through a channel (say, a marketing email), the CDP should remove them from being targeted by another channel (say, Facebook ads); or, through the unification of data can begin to build channel propensity models to optimize engagement based on where they are most likely to respond (e.g. never clicks ads but opens push notifications).
What to beware: Vendors that tout “machine-learning” or have opaque optimization offerings but do not offer bi-directional integrations with critical feedback loops.
5. Emphasis on data governance
Why should you care: Whether for privacy and regulatory reasons, data volume optimization or experimentation, you should have a range of methods available to gain control over your data flows. You should be able to set forwarding rules based on data types, values, identities, audience membership, and more.
What to beware of: CDP Vendors who do not have an extensive privacy framework for GDPR and other regulatory acts or at least APIs to support data deletion at the user level.
Some parting thoughts
While there may be overlapping features, a CDP is no replacement for a marketing automation solution (and vice versa), same with a demand-side platform, data warehouse, etc., and it’s not a silver bullet by any means. It requires proper investment and cross-functional buy-in rooted in a coherent data and identity strategy.
However, once you have a data strategy and cross-functional support, a CDP will add significant agility and resilience to your infrastructure and organization, as well as reduce your operating costs, plus provide the necessary controls for proper data governance.
It will be exciting to see the space continue to evolve.