DMP versus CDP: Which tech platform will win the marketing war?
- 29 July, 2019 07:18
It was just last month Salesforce officially took the wrappers off its first customer data platform (CDP).
The move follows in the footsteps of other enterprise martech players including Adobe, IBM and Oracle, who’ve all also debuted ‘CDP’ versions of their own. Initially a category full of pivoting tag management vendors and best-of-breed upstarts, the fact enterprise martech players are now seriously adopting the CDP approach suggests it’s one that’s here to stay.
Certainly CMO’s own State of the CMO research indicated the growing popularity of CDPs, with 16 per cent of respondents stating plans to purchase a CDP in the coming 12 months.
Which raises a big question: What does it mean for the longer-standing data management platform (DMP)? Will these platforms will continue to co-exist, or will CDPs in fact replace DMPs?
Salesforce itself is a big backer of the DMP, outlaying about US$700 million in 2016 to acquire leading DMP player, Krux. At the time of announcing its CDP, former Krux A/NZ leader and now Salesforce AVP of data and audiences, Jo Gaines, told CMO the decision was needed to adapt to the continually evolving customer.
“Back in the DMP early days, it was very much about helping customers to manage frequency, see where their money was being spent, what audiences were looking at campaigns. Now, this evolution is requiring access to all the data on a customer in the one place,” she said.
“It’s not about creating something you have to buy, it’s about connecting up everything you have and stitching it together. Here’s an easy way to use an identity layer that works across all of it.”
Pivoting versus innovating
Briefly, CDP describes platforms that create a persistent and unified customer database pulling in disparate data sources – and notably, first-party data - in order to present and then execute off a unified customer profile. It’s a term first coined in 2013 by David Raab, founder and principle of Raab & Associates, who’s since launched a CDP Institute to oversee the category’s rise and respectability.
Forrester senior analyst B2C marketing, Xiaofeng Wang, agrees CDP is a hot term gaining increasing amounts of attention. Yet she positions them as both immature and less disruptive than you may think.
“Most of the noise is from vendors who pivot into, rather than build, the category for the marketing benefits,” Wang tells CMO. “It lacks real market demand – only 15 per cent of inquiries Forrester received about CDP came from marketers.
“The reality is CDP is a loosely defined category, and still at very early stage.”
It’s for this reason Wang believes CDPs won’t replace DMPs or any existing marketing technology anytime soon, because the product features are still nascent.
“CDPs today lack functional parity with existing solutions. Early CDP adopters provided feedback like not effective or doesn’t do as well as expected,” she says.
CDP vs DMP: Technical capability
So what is it DMPs offer that CDPs can't, and vice versa? Wang notes firstly CDPs only cover insights of existing customers, while DMPs are actively used to execute against unknown prospects. This fundamental difference is described as PII vs non PII, or known customers vs unknown customers. CDPs store PII data, and DMPs store only anonymised data.
“CDPs offer a streamlined and consistent method for collecting, connecting, and activating first-party marketing data, but not so much and second- and third-party data. DMPs can do all three types of data,” Wang says.
Nevertheless, CDPs and DMPs have some overlaps in use cases they serve for marketers. These include audience analysis and targeting, basic personalisation, and some cross-channel campaign management capabilities.
However, CDPs don’t cover programmatic media buying, which is one of the key use cases of DMPs. Two other main DMP use cases are cross-channel marketing execution across digital channels, along with data-driven decision making in digital marketing.
According to Raab, DMPs and CDPs have often been confused but they are really quite different.
“DMPs are designed primarily to manage audiences for online advertising; as such, they deal mostly with anonymous data that becomes obsolete quickly [cookies] and often with third-party data not collected by the advertiser itself,” he explains. “Where DMPs support real-time bidding for programmatic ads, there is an additional requirement to return data very quickly – typically around 30 milliseconds. DMP designs and technology are optimised for these tasks, which involves storing relatively simple data.”
This is done via a collection of tags indicating individual attributes such as ‘male’ or ‘in market for a home’. DMPs also handle very large numbers of IDs to access quickly.
On the other hand, CDPs are optimised for different requirements. They primarily store first-party data and personal identifiers such as name, address and phone number; they store much greater detail such as the history of Web pages viewed; and they retain data indefinitely, Raab continues.
“CDPs do have real-time capabilities but of a different type - they return much more data and sometimes predictive model scores or recommendation - but they are slower, with a response time measured in seconds not milliseconds. This requires a different technical approach from DMPs,” he says.
In fact, Raab has built a five-point checklist for identifying CDPs. These are: Ingest data from all sources; retain full detail; store data indefinitely; build unified profiles; and share the data with any external system.
“Compare with DMPs capabilities, and you’ll immediately see DMPs falling short to some degree on every one, except possibly data sharing,” he says.
Which is why Raab anticipates a clear demarcation between the two categories long term.
“It’s very common for a CDP to feed customer profile data into a DMP and for a DMP to feed ad exposure and response data back to the DMP,” he says. “DMPs are much better at integrating with the rest of the advertising ecosystem and some extend to ad campaign management and analytics. CDPs are better at ingesting all types of data, storing it without loss of detail, and making it available in different formats as needed by different systems such as predictive model building, campaign management, Web site personalisation, and cross-channel attribution.
“The technical differences between the systems make it extremely difficult for one product to combine the functions of both. It’s not impossible but any claim to do this should be examined very closely to understand the actual capabilities being presented.”
As IDC marketing technology analyst, Gerry Murray, puts it, the difference between CDPs and DMPs is data, use cases and workloads.
“DMPs typically process anonymous data about ad clicks and Web visits and use it to try to improve the performance of various offers across channels, primarily in digital advertising,” he tells CMO.
“CDPs are mostly but not exclusively about processing first-party data with personally identifiable information and using analytics to do complex segmentation for outreach across channels including email, text, and ads. Both ingest and store data from other systems to reduce lag times between interactions and analysis.”
Read more: The lowdown on customer data platforms
While less forthright on DMPs, Murray is fairly confident they’ll remain in the martech ecosystem as a solution for brands with huge databases, traffic volume and big media spend.
“CDPs are more likely to be the basis for or at least an integral part of larger enterprise data integration efforts, especially for brands that want to advance their analytics and prepare to take best advantage of machine learning and AI,” he says. “Data preparation tends to be the biggest time sink for most AI projects and since it is usually a heavy lift that can take a year or longer in a large enterprise, there is a clear advantage to early movers.”
Raab agrees DMPs for first-party data are still primarily used by large enterprises with large numbers of customers and hefty media spend.
“DMPs with third-party data are accessed by smaller companies that need access to non-customers. Those DMPs might be hosted by an ad exchange or by a media company,” he says.
Yet confusion continues to exist. Raab attributes this to DMPs being sometimes over-sold as complete solutions to managing all customer data largely in Europe, but also the US.
“One result is a disillusionment with DMPs that obscures the value they do offer for their primary use of advertising audience management,” he says. “Some DMP vendors have attempted to reposition as CDPs to escape the bad reputation of DMP. This is hard because of the technical constraints – it requires substantial new engineering to switch from the DMP design to a CDP design.”
Despite this, several DMPs have moved partly in the direction, for example by adding features needed to securely manage personally identification information (PII), which DMPs traditionally excluded to avoid privacy issues, Raab says.
“But this is only one feature and doesn’t by itself give a DMP the capabilities needed for a CDP,” he adds.
In terms of DMP innovation, Xang notes rising utilisation of artificial intelligence (AI) should further unleash the power of data through advanced analytics. “Leading DMPs are investing aggressively in machine learning and data science to elevate capabilities such as predictive analytics,” she notes.
Raab sees AI helping DMPs do better targeting by finding look-alike matches between one set of profiles, such as current customers and others, such as non-customers.
“It can also support campaign optimisation by finding segments within an on-going campaign that are responding well or poorly and adjusting the targeting criteria appropriately,” she says.
With the ad stack largely automated, AI is now adding incremental efficiencies in terms of more nuanced bidding strategies, more granular audience segmentation, dynamic content optimisation, automated content tagging, continuous testing and optimisation and more, Murray says.
“AI is adding similar value on the CDP side so it is not in itself a key differentiator in the long run,” he says.
Whatever the trajectory of both platform approaches, it’s clear the industry is entering a future that requires an integrated data infrastructure with components optimised for specific workloads and use cases, Murray concludes.
“DMPs and CDPs and master data management, streaming analytics, and more will be required in many large organisations to handle the wide variety of data and volume demanded by different use cases and the governance policies that need to be applied to different parts of customer records,” he says.
“I’m not expecting one platform to do everything. I don’t think that’s technically nor practically ideal as different teams and their budgets will still determine which solutions are used for which data sets.”
Follow CMO on Twitter: @CMOAustralia, take part in the CMO conversation on LinkedIn: CMO ANZ, follow our regular updates via CMO Australia's Linkedin company page, or join us on Facebook: https://www.facebook.com/CMOAustralia.