What Is This Thing We Call A CDP?

Data-Driven Thinking” is written by members of the media community and contains fresh ideas on the digital revolution in media.

Today’s column is written by Martin Kihn, research vice president at Gartner.

A new technology appears, seemingly from the ether, and promises to change our lives. Customer data integration, labeling and storage problems will disappear. Identities will merge. You’ll be able to find new audiences until your ribs squeak and deliver them to any execution system in the barn.

Oh, and it will scale, rarely fail and enable (yes) true one-to-one marketing.

The name of this magic machine is … CRM. It was 1998. Companies piled in, dropping $3.5 billion a year on apps and databases alone – consulting fees not included – and yet, by 2001, 50% of CRM projects “failed.”

The same thing happened, on a lesser scale, during the great marketing automation boom of the 2000s. And it’s happening again.

Marketers unite! You have nothing to lose

It takes a generation to forget our parents’ wisdom. This fact is great for startups, whose founders and funders tend not to look back anyway. But it should inspire buyers to wonder what we can learn from the past.

No marketing tech in recent memory has shot up the hype curve so far, so fast, as the customer data platform (CDP). One new vendor received more RFIs last month than it had all last year. Companies with related solutions, such as tag management or data integration, are rebranding themselves as CDPs. A thriving Customer Data Platform Institute (CDPI) formed and already features 55 logos.

Big clouds smell a threat. VCs hear that party sound. Engineers get busy. And it is left to the hard-working marketer to wonder: What is a CDP, anyway?

The category was named in 2013 by David Raab, a respected mar tech analyst and founder of the CDPI. He put forward some sensible criteria, including a focus on first-party data and known identities and an ability to connect with common marketing systems for input and output.

Above all, the CDP has two guiding principles:

  1. It is a clean and open database for customer information
  2. It is owned and operated by marketers

Although it took a few years – some might say, until this year – for any real market to appear, those two principles were, in retrospect, the CDP’s Declaration of Independence from CRM and marketing automation.

The CDP is not the new CRM, because CRM is not a category. It is 190 different capabilities that span five areas – only one of which is marketing – and a massive cross-functional domain that scrambles from information and asset management to personalization and voice-of-customer.

And that’s the problem. CRM is labyrinthine, slow, cross-functional and controlled by IT. Marketing is fast, loose, self-centered and (if we’re honest) annoyed by IT. It wants its own flexible customer database. Now.

Marketing automation and multichannel campaign management systems still fill this need for many, but they’re showing seams. They are purpose-built, limited in scope and can have rigid taxonomies. They are obsessed by email and aren’t built for tag and API management, identity mapping, data integration – all things marketers say they want from CDPs.

The great CDP-vs.-DMP debate

It is ironic that the technology most often confused with the CDP is the one it resembles least: the data management platform (DMP). As Luma Partners’ Brian Andersen pointed out in his Industry Preview closing keynote this year, CDP vs. DMP is “not an or … it’s an and.” They do different things.

The DMP negotiates our beloved programmatic advertising, while the CDP – by definition – is grounded in individuals known by name, email, customer number or another personal ID. The DMP operates on massive audiences; the CDP, on a sensible number of souls. When Salesforce acquired Krux, a DMP, it encountered a data store many times larger than all its clouds’ previous requirements combined.

And the DMP and CDP are complementary, spanning the customer journey from ignorance to devotion and winback. They have much to learn from one another and a similar mission, namely personalization. Someday, they may be artfully combined.

In fact, the technology CDPs most resemble are tag management systems. It’s no coincidence that major tag management systems are now CDPs and that at least two well-known CDPs started as the mobile equivalent of tag management systems. The intent of the tag management systems was to zigzag IT and give marketers control over data collection; add a database and identity management, and you’ve got a CDP.

It is also no coincidence that right around the time marketers started chanting five-syllable words at conferences – probabilistic and deterministic – the CDP appeared. Identity is required for people-based marketing, and a customer database that can’t match devices and browsers and so on to people isn’t better than the status quo.

Naturally, most CDPs claim to provide identity management. Examine this claim. Identity management traditionally meant fuzzy matching on personal info and using tables to link records to a common ID. But most marketers today think of “identity” as a synonym for the cross-device conundrum.

But solving cross-device requires something that CDPs do not have: an identity graph. Graphs require alliances, co-ops, hard math and time. There is a reason people still need companies like Acxiom and Neustar.

Back to the future, part deux

Admitting it’s a new space with the requisite froth, what can the marketer do to set realistic hopes? How can we miss that CRM Y2K moment?

First, know what this product does. Last month, the CDPI issued a good vendor definition for public comment. Tempering this with my own Irish skepticism, I think the CDP can be pictured like this:

Click here for larger image.

It facilitates data collection, unification around a persistent ID, flexible storage and easy access from outside. You’ll notice that the areas of “predict” and “decide” – the decision-making parts of the tool – are labeled “optional.” They’re optional for CDPs, but should not be for marketers.

Be careful here. It’s a region rife with false whispers and future broken hearts. It’s where the dreaded shibboleth “AI” appears in pitches – where most CDPs have done the least amount of work – and where the greatest future differentiation will occur. Everything else is just transportation.

At the height of the CRM boom, Gallup issued a moving screed called “Is CRM All Hype?” It suggested some principles we can adapt to our CDP moment. Combined with my own observations, they are:

  1. CDP is not a system of record – it’s a system of innovation
  2. Failures won’t be caused by the tech – they’ll be caused by you
  3. You probably don’t know how messy your data is now
  4. You will overestimate your team’s technical skill
  5. You want short-term results but use long-term evaluation metrics

This last point is the one stressed by Gallup. What does it mean? Simply that most marketers say they want to improve the customer’s experience while caring only about their own experience. They’ll use the CDP to boost quarterly revenue but deem it a failure when retention falls.

There is no CRM system – and no CDP – that can change the existential truth that the needs of the business and customer are almost always in tension.

Follow Martin Kihn (@martykihn), Gartner (@Gartner_inc) and AdExchanger (@adexchanger) on Twitter.

Enjoying this content?

Sign up to be an AdExchanger Member today and get unlimited access to articles like this, plus proprietary data and research, conference discounts, on-demand access to event content, and more!

Join Today!


  1. Martin,

    Thanks for sharing an informed perspective. Agree with the principals that you call out.
    Failures won’t be caused by the tech: A CDP (like all MarTech) enables customer-centric strategies. Technology won’t overcome the wrong strategy.
    You will overestimate your team’s technical skill: I will go further that a successful CDP implementation requires more than just technical skills. It requires vision and a new level of collaboration across teams. You likely need a change agent that is supported by leadership. Don’t understand the human element of a CDP implementation.
    You want short-term results, but use long-term evaluation metrics: Agree that CDP’s should not be viewed through a short-term lens. Ideally you will identify near-term opportunities, but be prepared to play the long game.

  2. Martin – great article!

    I’m curious about your point that “CDP is not a system of record – it’s a system of innovation”. Do you believe that there is a system of record in the martech/adtech stack? From my perspective, a CDP can be a system of record for certain categories of companies e.g. digital natives, but probably not for others e.g. CPG or Auto.

    In parallel, one trend that will be interesting to watch is how CDPs and DMPs combine forces or capabilities to create a true data system of record.

    • Mike, I think a CDP most certainly “could” be utilized as a system of record but at Martin points out, that’s not really it’s purpose. It’s supposed to be a system of innovation. Which is a rather obscure definition but I think that’s on purpose as the system has many potential use cases.

      The CDP was designed to give the power of data analysis, segmentation and distribution back to the marketer. Anyone who’s worked with large brands and their legacy CRM/Data platforms knows the pain involved in segmenting and distributing data. It’s an incredibly arduous process, which can sometimes take days/weeks or even longer.

      This makes it incredibly difficult for marketing organizations to actually utilize their incredibly important and powerful CRM data. Enter the CDP, which can sit between database and execution platforms to provide insights, create segments and then distribute them, in near real time. Meaning marketers can finally design and run complex experiments and get actionable insights and data back in time for them to actually do something with it…

      I think the system of record still needs to exist. It has to be highly scalable, super secure and support very complex custom integrations. The CDP needs to be able to extract what is important for a marketer without having the burden of also being the system of record.

      Just my $.02….

  3. Martin, this is a fantastic summary. I’d like to add some context based on my experience AS a marketer who also happens to be working for a CDP company.

    Short Term Value + Long Term Results
    I agree that the benefits of a CDP are best measured long term, but also think CDPs can provide short term value. To be fair, CDPs require more “heavy lifting” to achieve short term results (say, for real-time audience-segmentation for web personalization) but, you’re also building a foundation that can be applied to other customer data problems from around the Company and provide scalable, long-term value.

    CDP is Not a System of Record – Or is It?
    Well, if your CDP includes storage then you could argue it can be used as a system of record or “single source of truth”, but, I think one of the key elements of a CDP is that it compliments existing systems of records vs ripping+replacing them. What this means is that a CDP is a actually a “dynamic” system of record which automatically updates based on new information from the systems it’s connected to. This is important because, each line of business likely already has it’s own “system of record” and they should continue to maintain it. When they do, it should effect the end-results coming out of the CDP. So, for example, if Finance updates their Customer Lifetime Value number in their records, that can dynamically update programmatic advertising bids, or simple campaign ROI dashboards maintained by marketing. All without either team having to actively coordinate with each other.

    AI and ML
    Ok, those are two of the buzziest terms around so a lot of vendors talk about them. My interpretation of how a CDP fits into all of that is this: a CDP’s main function should be to solve the underlying data-set problem that FEEDS applications powered by AI/ML. The reality is that many “point” solutions have great algorithms that are purpose built to optimize their key features (ad optimization, churn prediction, etc) but if that algorithm only has access to a partial data set, than its results will be suspect at best. A CDP can solve for this by creating a complete and dynamic data set that can be rapidly iterated on to test different “experiments”.

    Technical Skill
    This is a funny one and, I would argue, is inversely proportional to the power of the CDP. There are some CDPs that are pretty easy to use, but their functionality is frequently limited to marketing and sales applications (which is a good start, to be sure). That’s probably enough for most SMB or some MM companies. However, larger Mid-Market and Enterprise companies will have home-grown, legacy and customized system they want to integrate with and a CDP that can do this will require an operator with a higher degree of technical skill or a vendor that can automate data-import. This is important because data import is a NON-TRIVIAL part of the process many CDP vendors gloss over as “engineering’s problem”.

  4. Marty, this couldn’t be a more timely blog about such a critical topic for marketers – especially those in large enterprises that have a certain amount of scale and complexity. Marketing is a undergoing a renaissance / upheaval / reckoning…you name it – driven by this explosion of customer data across an ever-increasingly complex journey…and the commensurate inability for (even best-in-class) marketers to use it effectively.

    I’d offer that the ultimate promise of a Customer Data Platform lies in the unrealized promises of the Martech point-solutions that litters most marketing orgs today, and the frustration stemming from their knowledge that their customer data is scattered and hidden right under their noses.

    Only with a complete picture of a customer in one place – across any data source, across any time period, matched to a customer ID of some kind and accessible – oh, and playing nice with all current Martech – can marketing actually execute personalization at scale.

    The other reality I think that marketing teams needs to come to terms with is that the days of bypassing IT are over. A CDP needs to be controlled by the marketing, but IT must be brought in early and reassured – especially in a GDPR world.

  5. Great post! Disruption is in the air again. Not all CDPs are created equal, and just as we saw ESPs aspiring towards a brighter promise of multi-channel, many tag management, DMP, and web personalization vendors aspire to be CDPs. The devil is in the (data) details and this system is not just for marketers and certainly goes beyond anonymous digital profiles. The opportunity I see for CDPs is not to be a single database, but to be a platform to access data from any location and format, cleanse and format the data, and create an accurate and complete customer view that is actionable for use in engagement across the enterprise. Data security, privacy, and governance are also crucial in this world and the name of the game is being able to leverage best of breed data and execution systems in an open ecosystem.

  6. How does this definition of a CDP differ from the mature category of real-time interaction management (RTIM)?