Customer Data Unification
What’s the difference between a CDP and a DMP?
If you read a high level description of a CDP and a DMP, you might think they perform the same functions. They both build unified, segmentable profiles and send data out to other systems.
But not all profiles are created equal. By looking closely at the nature of the profiles that each of these platforms build and how they are used, it becomes clear that each platform has distinct value and utility.
CDP: Known Customer Profile Builder
CDP = Customer Data Platform
The primary job of a CDP is to unify a brand’s first-party data to build unified customer profiles using known customer data. Data sources include eCommerce transactions, point of sale transactions, CRM data, clickstream data, loyalty databases, email response data, social interaction data, and any other dataset that contains customer data.
The profiles a CDP builds contain PII (personally identifiable information) such as known individual’s email addresses, mailing addresses, full names, birthdates, purchases and other transaction data, social and email interactions, preferences, and more.
This data describes how customers experience and interact with the brand itself, not the world at large. Therefore, it enables brands to know their own customers more intimately and build meaningful relationships through targeted, personalized interactions across online and offline channels.
Below is a visualization of a fictional CDP customer profile built from a brand’s first-party data about a specific, known customer. Note the PII, the derived fields such as customer lifetime value, the linked transaction data from eCommerce and point of sale systems, and data from several other first-party sources.
Because the data in a CDP profile contains rich, cross-source data and PII about known customers, it can be used in a number of meaningful ways. The two main categories of use cases that CDPs unlock for brands are:
- The ability to deeply understand their own customer base and power insights that drive the business forward in meaningful ways.
- The ability to drive effective, personalized, omnichannel marketing and customer experiences through the site, app, email, social, custom ad targeting, in-store experiences, call center interactions, and more.
More specifically, CDP profiles can be sliced and diced according to any of the attributes they contain and used to power micro-segmented emails to customers, site personalization using historical transactions, and custom ad audiences on social media (where email addresses or other personally identifiable information are required).
For example, brands could use a CDP to create segments of all their customers who purchased products in-store last month, who are highly responsive on both email and social, with a lifetime value of higher than $500, and who have a history of buying holiday gifts each November. The segment could then be used to fuel an omnichannel campaign via email and social media with personalized offers for specific holidays gifts.
DMP: Anonymous Consumer Profile Builder
DMP = Data Management Platform
In contrast, DMPs build anonymous consumer profiles. They do this by capturing anonymous clickstream data from visitors via tags on the brand’s site. DMPs also take in a offline first party data, but it must be anonymized, i.e. hashed, or de-identified in some way.
This anonymization is vitally important. If profiles in a DMP contained PII, it would be illegal to match them to other anonymous attributes because of strict privacy laws and penalties. And this matching offering is core to the value proposition of the DMP.
The DMP’s proprietary data asset contains demographic data and information about people’s interests and shopping behaviors. Some of this information has been inferred using third-party browser cookies and their associated user behavior across the internet and a majority comes from 3rd party data providers. DMPs match a brand’s anonymized customer data to this data asset to form consumer profiles.
Below is an actual DMP consumer profile. Again, DMP profiles are anonymous and unified around users’ web browsers and computers. They contain no personally identifiable information.
Here are some of the attributes listed in the actual DMP profile above:
Marital Status: Single
Marital Status: Married
Female Head of Household
Male Head of Household
Net Worth: $25,000-$49,999
Net Worth: $50,000-$74,999
Net Worth: $75,000-$99,999
Net Worth: $100,000-$149,999
Net Worth: $150,000-$250,000
Net Worth: $500,000-$749,999
Net Worth: $750,000-$999,999
This profile tells a brand that the person who uses this browser is either male or female, married or single, and has a net worth somewhere between $25K and $999K. This lack of accuracy is typical for DMP profiles because the majority of the data has been inferred based on browsing behavior.
DMPs enable targeted advertising and site personalization. When a brand wants to show ads to customers before they have much first party data about that person or personalize a website, a DMP fills the information gaps with demographic and interests data, so people can be targeted with advertisements and content that are more likely to be relevant to them. They do this by enabling brands to create user audiences.
The trouble is, if a brand uses their DMP to create an anonymous audience of profiles that are believed to belong to married, female heads of household with net worths of $999K, the profile depicted above will be included. If the brand uses their DMP to create a different segment of profiles that are believed to belong to single, male heads of household with net worths of $25K, this profile will be included again.
Focusing on advertising, the result of these errors at scale is that many people will be shown ads that are not relevant to them. This is costly for brands, resulting in lower ROAS. But it has historically been considered low risk because ads are just, well, ads. We have all been subjected to innumerable ads that have no relevance to us. Over time though, as internet-first brands like Netflix and YouTube continue to raise the bar on relevancy, this status quo could shift.
Finally, unlike CDPs, which can fuel any external system with customer data, DMPs only connect with sites and DSPs for media buys against the anonymous segments they produce. This means DMPs have limited use cases.
CDPs build known customer profiles rich with PII, transactions, and interactions with the brand itself. These profiles enable in-depth analytics and personalization across channels, and are useful for building long term customer loyalty, maximizing campaign performance, and informing core business strategies.
DMPs produce anonymous consumer profiles with demographic and interests data unified around web browsers and associated behaviors. These profiles enable targeted advertising with varying degrees of accuracy, and are most useful during the initial period when a brand knows very little about a given customer.
CDPs and DMPs have mostly stayed out of one another’s swim lanes, but lately, more brands are bypassing DMPs and sending customer data directly to social media platforms for targeted advertising.
Instead of using the anonymous profiles that DMPs offer, they are using their own PII data to match directly to Facebook’s profiles to create targeted audiences via lookalike modeling and custom audience tools. Facebook profiles are also rich in demographic attributes, but also contain treasure troves of interests, preferences, and intent-to-purchase data.
In my next post, I’ll compare the path from a CDP to a DMP to Facebook against sending data directly from a CDP to Facebook, to see which approach drives better results. Interested in learning more about an Intelligent CDP? Check out our post on Intelligent CDPs here.