Evolution of CDP – Customer data platform
Conversion rates on e-commerce websites rise an average 15 – 20 % and engagement rates rise 30 % after implementing better personalization. Personalized emails have at least a 6 % higher open rate.
Defining concepts like “personalization” and “1:1” is not as easy as it seems. But it is clear from the data that at a basic level, personalization requires different brand channels to behave in concert.
Loyalty to consumer brands continues to decline, as the importance of in-the-moment experiences outweighs brand equity.
We know that customers require unprecedented levels of personalized experience from all brands and are voting with their wallets if they don’t get it. At the same time, these same customers are increasingly wary of providing access to the very information that is required to provide that experience–namely, behavioral, attitudinal and demographic data at the individual level. It’s a tension that researchers increasingly and aptly refer to as “The Privacy Paradox”.
Our recent research showed that the number of significant data sources used by marketers alone grew 50 % from eight in 2019 to twelve (projected) in 2021.
We believe the new paradigm of Customer Data Platforms is an evolution of what you’ve done before and is compatible with existing solutions. Our approach is one of compatibility and growth, not replacement and self-imposed crisis.
CDP – customer data platform
We will argue that CDP is just the latest evolution of the CRM category, with an emphasis on marketing use cases to begin, quickly expanding to other areas such as service, sales and commerce. We will define the key components of the CDP, from data ingestion, processing and identity management; to segmentation, machine learning and artificial intelligence (AI); to cross-channel activation, reporting and optimization.
We argue that a true enterprise-grade CDP must provide:
- Anonymous to known. Since customer journeys usually start with an anonymous ad viewer.
- Insights and engagement.
- System of engagement. Channel optimization, next- best – offer management and dynamic creative optimization.
- System of insight: “single view of the customer”.
What’s keeping marketers from achieving the “right customer, right message, right time”. The problem is not the amount of data being generated, but where that data is stored and who is using it.
“Known” data is any type of customer data that is personally identifiable, called “PII” or “personally identifiable information”. The CRM system is the operating system for customer data. The service system in the call center. Marketing and commerce systems.
Customer data is the lifeblood pumping through those systems and powering their operations and, over time, those systems have evolved to leverage customer data in the service of different outcomes: revenue (sales), customer satisfaction (service), direct purchase (commerce) and engagement (marketing).
Breaking down those silos requires companies to resolve customer data and create data portability across applications.
Customer resolution is the method by which companies create a data model to standardize those fields, and also build the ability to make those systems smarter by applying logic to map fields together accurately. Once a model is established for resolving customer data, the next step is to make different attributes available to other systems or “data portability”.
The typical customer owns 4.6 connected devices, has hundreds of active cookies and different IDs for every platform she interacts with (an Apple ID, Google ID, unique device ID for Xbox, etc.).
Whether you are a data-rich or data-poor marketer, the consumer’s mandate is for you to use their data responsibly to provision an experience across channels that acknowledges you know and understand their needs.
Cross – device identity management (CDIM) is the discipline behind making sure Joe is recognized as a person, rather than a collection of IDs.
Today’s complicated data privacy landscape makes applying such data harder than ever and that is another use of DMPs: managing consumer consent at highly granular levels and applying it at scale to ensure advertising is compliant with consumer rights.
Different consent flags included data collection (capture your device or cookie ID), analytics (include your browsing behavior or other attributes in reporting), targeting (use your data to show you an ad or experience), cross – device (include your ID within a graph of other IDs representing different device types), data sharing (the consent to share your ID with another party), and reidentification (the ability to associate your pseudonymous ID with a PII-based ID to customize known messaging).
Perhaps the biggest problem to solve in marketing over the last 15 – 20 years has been that of “identity”.
It has spawned an entire industry called “data onboarding” which offers the ability to put email addresses into a system and turn them into cookies or device IDs on the other end.
Many companies claim to be data-driven, but a look under the hood quickly reveals the insidious nature of siloed customer data. Although it’s clear that customer data is most powerful when completely unified, it tends to organize itself around specific business functions, leading to more silos.
At its heart, the promise of the CDP is to go beyond solving for disparate data, and start to solve for unified business processes, putting customers at the center of every decision.
The only real way to break down the people silos that prevent cross-team collaboration is to unify the data completely and start to build incentives for every stakeholder to enrich the core people data asset.
In many ways, CDPs represent an evolution of DMPs.
Firms like Adobe, Oracle and Salesforce to be the eventual leaders in the CDP realm – but a majority of our respondents believe that they already are de facto CDP providers.
From our studies we see these use cases as priorities:
- Predictive marketing and advertising;
- customer profile management and expansion;
- customer segmentation;
- development or enhancement of marketing automation systems;
- and content delivery systems that reach the right person at the appropriate stage of their customer lifecycle.
Customer resolution is the strategy for provisioning a single, unified ID or “golden record” that can power the new customer journey. Data portability is required for connected unified customer data across applications.
Matching known data to unknown IDs (data onboarding) and leveraging contextual and behavioral signals gathered pseudonymously to enrich known data (reidentification) are the two main methods for connecting these different sets of data.
History of CDP
The modern CDP emerged as a natural evolution of customer relationship management (CRM) and relational databases.
Already, direct mail is a prototype of the modern customer data-driven channel; we are still trying to reproduce it in the digital space.
Direct mail’s online cousin, email, was invented in the late 1970s, but it took the coming of the consumer internet in the 1990s to make it a marketing workhorse.
The challenge with storing data in tables is obvious to anyone who spends time thinking about it, and was obvious to data admins from the beginning: tables can get very, very large and complex very, very fast.
An answer to this problem was proposed by an IBM computer scientist named E. F. Codd in 1970 and was called the relational data model, which led to the invention of the relational database. The innovation Codd proposed was to relate many smaller tables to one another in a structured way, using a common key or data point, so that complex information could be broken down into a series of much simpler, interwoven tables.
A new challenge arises, however: how can the marketer do analyses on the multiple relational data tables? The answer is by using Structured Query Language (SQL), a query system designed to work directly on relational databases. Database marketing became a profession, and it consisted primarily of writing SQL queries to produce customer lists based on criteria described manually to match a particular campaign.
Systems designed to automate business functions related to customers and accounts started to appear in the 1990s. The umbrella term still used for such customer-facing automation systems is customer relationship management (CRM) software. In fact, at last count, CRM encompassed 190 categories, with many large organizations employing hundreds or thousands of instances.
One key subset of CRM is marketing automation. Marketing automation introduced a new database – specifically, for campaign-level data. Customer lists were renamed campaigns and these campaigns could easily be automated.
The result was a natural process of accumulation, both of tools and teams, and the brand ended up with what its Chief Digital Officer not-so-lovingly described as “a train wreck”. There were at least four uncoordinated teams, and three different agencies (creative, media planning, and CRM), each aligned around a different channel.
We’ve seen how the internet technology landscape of a typical brand became overcomplicated and inefficient, as channels piled on channels and teams spun up next to other teams.
The team decided that there needed to be both a place to store customer data and do analytics and another layer that takes an abstract of the customer data and combines it with real-time systems to do tasks like real-time interaction management, decision and content assembly.
What would an idealized marketing architecture look like, anyway?
It would require:
- Unified user profile
- Smart segments
- Plan and react capability
It’s the story of the martech space in general. It was growing steadily until about 2014 and then took a hockey-stick leap in the number of vendors and categories.
What CDPs try to do is create an environment where there is a more flexible data model you don’t have to define in advance.
But putting aside the label of CDP, it’s hard not to see a data platform as the bedrock of the martech stack. Everyone needs a flexible model for data around the customer.
The ideal marketecture : Companies may not be able to rip and replace their Frankenstack of legacy systems, but they can strive to re-architect their systems within a framework that focuses on: a unified customer profile, intelligent segmentation, rules-based journey management that leverages AI for decisions, API-driven engagement to every endpoint the consumer interacts with and optimization that works within a closed loop where historical data feeds go-forward recommendations. This is the recipe for success.
What is CDP – customer data platform
Every company wants a CDP, but for slightly different reasons. The category is new, largely undefined and confusing. As an example, in 2019, research firm Advertiser Perceptions asked hundreds of senior-level brand marketers and agencies, “What CDP have you used in the past 12 months?” The top three answers were Salesforce (62 %), Adobe (49 %) and Oracle (42 %). The problem? None of those companies had an actual customer data platform in the market at that time!
CDP as a martech category was born in 2013, in a series of blog posts and a report written by an independent martech analyst based in Philadelphia, named David Raab.
Gradually, a set of capabilities loosely coupled under the label “Customer Data Platform” emerged. This happened as a theoretical exercise, since no single vendor yet existed who was capable of delivering this vision (no matter what they said). The key components are:
- Acquire: The ability to natively collect and extract data from a wide variety of common sources.
- Process: Built-in tools to cleanse and manage both attributes (fields) and identity markers (IDs).
- Expose: Make the normalized, ID – mapped data available in a persistent data store.
- Analytics and decisions: Execute decisions and perform predictive analytics on the data.
- Delivery: Once decisions are made, send the signals to the engagement systems that actually deliver the message.
From the beginning, CDPs were more about data management than built-in analytics.
In other words, the promises made by CRM for marketing systems in the past are almost identical to those made by CDP for marketing systems today.
A healthy cross-section of marketers told us they wanted a tool to help them with the following issues:
- Cross-channel campaign management
In order to achieve the outcomes above, marketers told us they wanted a tool to help them do the following key tasks:
- Profile building
- Data ingestion
- Identity management
Every CDP will do some things better than others, but every offering that wants to claim membership in the category will need these basic capabilities: data collection, data management, profile unification, segmentation and activation and insights.
A good CDP will have data connectors built in, to enable inbound data partners to plug in and also be able to capture data from both known (PII-based) and unknown (pseudonymous) sources.
When we think of “data management” specific to CDP, we are talking about the classification of data within the system, how it is defined, and the overall schema the data is organized into.
The basic idea of profile unification is to move from static, singular “contacts” to rich profiles of people that not only unify multiple contact records for the same person, but also include the other identifiers they use (browser cookies, device IDs, a company’s first-party ID, etc.).
Profile unification depends on the two capabilities above (the ability to capture data from different sources, and the ability to “manage” it into a common taxonomy). Profile unification lies at the very center of CDP. You cannot do it well without data collection and management. Once all the relevant data has been associated into a unified profile, marketers need a way to organize customer data into scalable segments to get reach.
The promise of CDPs is not only aligning customer data – but unifying the place where marketers do the work of segmentation. Once segmentation is complete, activation is the next priority.
Insights. A good CDP will initially focus on data-specific analytics, looking at things like consent across the data set, migration of users in and out of segments, broad cross-channel engagement, lifetime value (LTV) and propensity scoring.
Companies wanted a “golden record” of the customer – a profile that would serve as the single source of truth for customer data that every stakeholder across the enterprise could refer to. We started to think about this system of data organization as a “system of insight” – more concerned with what one could learn about a customer than how one could interact with them. These “systems of insight” would serve as the single source of customer truth, the aggregation point for the many different sources of organizational data.
Other proposal requests featured quite different use cases, centered around using data to personalize interactions customers have across the typical customer journey. These systems were required to have data in a real-time customer profile store, red hot and ready to be accessed by other applications to make decisions in milliseconds. Unlike the “Insights” CDP, these systems favor actionability over the manipulation of customer data for segmentation.
What if you could put the best of both worlds together?
What are the requirements for building such a system? We think there are three and they set a high bar for success.
- Known and Unknown (CDMP) Data Must Be Unified – Customers tend to weave between “known” channels like email and “unknown” channels like a website at will.
- A Business-User Friendly UI – Unlocking the value of unified customer data requires an interface that enables business users to query, analyze and segment customer data on the fly.
- A Platform Ecosystem The Third – arguably most important, aspect is the way in which tomorrow’s enterprise CDP will be built. The marketing world is moving from “suites” with lots of loosely connected applications to “platform ecosystems” which can be provisioned on a “common core” infrastructure.
In some enlightened quarters, David Raab is known as the “Godfather” of the CDP. CDP as a software category was born in 2013 to address “the integration gap” in data that created silos between marketing channels and the groups in the organization that managed them.
Organizing Customer Data
At the highest level, the key steps for successful data management are:
- Acquire: Data collection
- Organize: The process of transforming data from disparate original sources into a common data model
- Analyze: Analytical processes
- Deliver: Sending signals to systems of engagement with instructions and content, based on embedded decisions
The purpose of the data ingestion or acquisition step is to identify data sources, ensure that the needed data is collected and produce it in such a way that it is available for the next steps.
The two main patterns for data collection are: Batch processing and stream processing.
What is harmonization in practice? The best way to think about it is as a way of mapping fields in different customer data sources to one another, when those fields are supposed to represent the same type of data. One best practice for data harmonization is to start with a standard data model, customize it for your particular requirements and use that as a canonical model or template for your final format.
Managing identity can power more effective multichannel engagement, improve targeting and creative tactics, and make measurement more accurate. Specific benefits of identity management are: Improved media effectiveness, more relevant content, better analytics and insights, more informed governance.
Identity management is built around what are known as keys. These are the ID handles that represent the specific proxy or known individual (or account). Common keys include browser cookies, hashed emails (HEMs), IP addresses, plaintext emails, and customer IDs.
Segmentation is used for a variety of purposes, including product development, pricing, targeting, messaging, and measurement. We believe the most common segmentation models among marketers tend to be based on:
- Customer value
- Product preferences
- Price sensitivity
Attributes come in three basic types: they describe what a person (or account) is (demographics/firmographics), what they do (behaviors) or what they think (attitudes).
Fast, accurate segmentation based on real-time profiles may turn out to be the “killer app” of the category.
Customer data activation is a broad discipline that could be seen as encompassing pretty much everything a marketer needs to do.
First-Party Data and Privacy
Web browser cookies have been around for a long time. They were invented by a 24-year-old Netscape engineer named Lou Montulli in 1994.
We’ve entered an era of avoidant consumers, who are increasingly resistant to data collection and use by companies and their marketing departments.
Digital data privacy is maintained by four parties with (at times) different agendas. These are:
- Web browsers and standards bodies
- Government regulators
The so-called “Seven Principles” of GDPR are:
- Purpose limitation
- Data minimization
- Storage limitation
At a practical level, we approach privacy and compliance as a three-tiered practice:
- Classify data
- Determine user entitlements
- Ensure compliance
The message is clear: a healthy majority of consumers are willing to share data in return for relevance – as long as it is collected and used in a way, and under structural controls, that makes them comfortable.
When we’re asked if we value our personal data, almost all of us say “yes”. Yet our behaviors show otherwise. The trouble with “rational choice theory” as it’s called, is that we’re usually forced to make decisions without enough information.
Marketers and advertisers are going to have to master the art of gaining consumer trust. Don’t talk about people behind their backs. Give a sense of control. Explain the benefits in concrete, positive terms. Remember, people are different.
Customer-Driven Marketing Machine
Know your customer as much as you can through collecting data; personalize every interaction where the customer encounters your brand; and engage across at the channels and surfaces they touch. Later, we added a piece: measure everything and use that data to get better every time you start a new journey.
- The “know” was data management.
- The “personalize ” was Artificial Intelligence.
- The “engage” was journey management.
We think of the engage layer as the ability to provision an experience across channels in two distinct modalities. First, the ability to manage “customer journeys” in a prescriptive way. Second, the ability to provision a real-time customer journey “in the moment” as the engagement is happening. Journey management is the ability to orchestrate the way a customer engages with a brand across a number of channels over time. Real-time Interaction Management (RTIM). RTIM is the “in the moment” version of orchestration, requiring different tools and more real-time data capabilities.
We think there is a fairly simple working model or philosophy, which can help structure a CDP project.
- Team – Every transformation comes together around “people, process and technology” and we argue that, while you need all three, the people are probably the most important aspect.
- Platform The “platform” is the investment, but it is also an operating philosophy.
- If CDP is the tool at the core of “transformation”, it must democratize data across the organization and enable access to them across roles.
- Use Cases This may seem beyond obvious, but data unification for its own sake doesn’t produce value.
- You cannot track the effectiveness of your data transformation without a clear performance framework, and having KPIs to align with is paramount.
- Methodology A lot has been written about the wisdom of Silicon Valley firms that have the “permission to fail fast and fail often”, and much more has been written about why that might be a terrible idea.
- We’re talking about supporting a culture that allows for inquiry, allows for fast results through testing and rewards data – driven insights.
- Operating Model. Finally, the CDP project needs its own operating model to fund ongoing executive support and investment.
There are three primary stakeholders needed to get CDP off the ground the right way and many more in supporting roles. The primary three are Marketing, IT and Analytics.
In a CDP center of excellence, marketing should be thought of as responsible for the delivery of experiences that can be optimized through the use of customer data. The evolution of CRM over the years has moved customer data closer to the marketing consumer as systems become less technical and more accessible to business users.
The IT team must bless the CDP as the system of record for capturing, storing and unifying customer data. This means migrating data traditionally stored in data server farms, data warehouses and data lakes into a central store.
While the IT/CRM team may be the stewards of customer data and the marketing team the prime activator of them, the analytics team must own the segmentation strategy and set the metrics for success by building models that can measure the value that customer data brings to the organization.
The analytics team must also be responsible for helping create marketing’s overall ROI model and work to establish individual KPIs that can determine the value of data-driven marketing. This helps the COE determine how much value, by channel, can be extracted by adding additional segments and targetable attributes – and when precision-based methods start to yield negative results through diminished scale and added complexity.
When executed well, the center of excellence produces a cyclical, flywheel effect.
Better data leads to better segmentation. Better segmentation leads to better engagement. Better engagement leads to better data. Rinse and repeat.
The maturity scale progresses along two distinct axes: the ability to coordinate different marketing channels and the ability to personalize interactions and deliver relevance.
Channel Coordination Stages:
Engagement Maturity Stages:
Touchpoint marketing is oriented around specific campaigns. This is the way marketing has worked forever, and brands can be highly successful in this phase and never progress to the second phase of maturity: Journey- based marketing.
The key from moving from touchpoints to journeys has everything to do with a company’s data management capabilities (tie people data together) and technical ability to orchestrate (tie channels together).
The natural evolution beyond journeys is to move to “experiences”. It requires moving the marketing department’s orientation from channel-based interactions to being able to deliver the right message, creative, or offer in real time – “in the moment”.
Evolution is a move from systems of engagement to systems of intelligence. This is an exponential leap into AI-based marketing.
Adtech and the Data Management Platform
Quietly, in the background there was a key piece of technology capturing, unifying and helping to activate and analyze all of the little cookies and device IDs that powered the ad tech landscape: the DMP.
Big publishers harnessed the power of DMPs first. Advertisers soon followed, realizing they could use the same mechanics to gather consumer data, and also scour a gigantic pool of other people’s cookies and device data (third-party data) through marketplaces like Nielsen’s Exelate and Oracle’s BlueKai.
With enough device data (your IP address, device type, browsing behavior and location), you can be digitally fingerprinted and re-identified as a person. You are never really anonymous on the internet.
Beyond just helping both sides of the advertising equation manage the common online currency of people (primarily cookies), DMPs powered a great deal of functionality. They are audience planning, data activation, personalization, campaign optimization and delivering insights.
We think DMPs and CDPs are on a crash course and “CDMP” capabilities are required for total, “enterprise-strength” data management. This enables a single source of truth for customer data (both known and unknown), a single governance platform for managing channel-based consent, a single user interface for handling segmentation and a single place for activating data. This “CDMP” vision is not just a concept.
Today’s customers demand the seamless experience across touchpoints that go beyond marketing. This is what we mean when we talk about the “rise of the chief digital officer”.
Today, the most mission-critical job in marketing is working to increase the ability of the enterprise to identify customers and connect their data. Dramatic leaps in connecting the call center to traditional marketing are another reason CDPs are on the rise.
Another key area for enriching customer data is within e-commerce. Unlike Service, we’ve watched marketers integrate e-commerce (and in-store purchase) data into their marketing efforts for years. This begins with retargeting, continues into using historical purchase data for modeling and naturally evolves into advanced targeting methodologies around category preferences and interests.
By combining the rich, highly granular attributes of unknown visitors with known deterministic customer and purchase data can yield effective predictive models.
Ultimately, the successful data transformation strategy must look at the purchase not as the end of the journey down the funnel, but the seed that can grow into a new top-of-funnel campaign.
Every enterprise system that has taken a dominant market position recently has been a “source of truth” for a key business function.
The first benefit of bringing B2B CRM data into a B2C-facing CDP effort is the people data that lives within it. The first step is to distill the many different “contacts” in the B2B database into rich “profiles”.
Another key way to use classic B2B CRM data in the CDP is to leverage the ways in which the data has been organized into behavioral attributes.
Machine Learning and Artificial Intelligence
Customer data sitting in a data warehouse – even organized and cleansed in a true enterprise-grade customer data platform – serves no real purpose. It must be activated to yield business value. Data science is a hybrid discipline that draws from statistics, mathematics and computer science.
Typical tasks for a marketing data scientist include:
- Predictive modeling
All good marketing data scientists possess three key attributes: Quantitative skills; some knowledge of their industry; curiosity.
Algorithms used for machine learning build abstract representations of data called models. Labeled data is data for which the desired outcome is known and it is required to build predictive models. No model is perfect, but some are more useful than others.
The analyst must continually balance between the two extremes of memorizing the training data (known as overfitting) and building a weak model (known as underfitting).
Regression models use numerical data to make a numerical prediction of a desired value. Rather than predict a numerical value, categorical models attempt to classify or label something. Finding structure in numerical data generally involves creating groups or clusters that have some internal similarity while being relatively distinct from other groups.
Two of the most interesting ways machine learning has been used in marketing have been the areas of segmentation and attribution.
Just as there needs to be a single source of truth for people data, there must exist a single place to access that data for the purposes of segmentation. Machine-learned segmentation has been a DMP feature for some time. CDPs will take this AI application to the next level.
The advantage of living in a complex world is the availability of complex analysis and, if machine learning is good for anything, its finding statistically relevant deviations in large sets of data.
Two of the most exciting areas of current development in artificial intelligence for marketing are image recognition and natural language processing (NLP).
Most data lakes are not natively connected to many other marketing systems an enterprise owns – and getting modeled data to the endpoints of activation for advertising and marketing is an IT exercise that requires automating data ingestion, opening up infrequent “batch windows”, and aligning data lake customer data to the information models used by other systems.
In 1978, a man named Gary Thuerk used a precursor of the internet called Arpanet to message 400 contacts a promotion for DEC computers and created $13 million in sales from a single email message. The era of email marketing was born.
The role of the CDP in prescriptive journey building is to close the gap between the data needed to effectively deliver the experience.
If automating journeys is “touchpoint” marketing, then being able to deliver better journeys based on prediction is true “journey”-based marketing.
Leveraging AI-based prediction to provision journeys that cut across multiple channels is an art form that can increase conversion rates, reduce churn and tether previously disconnected touchpoints like service and commerce to the steel thread of a marketing campaign.
RTIM is defined as “enterprise marketing technology that delivers contextually relevant experiences, value and utility at the appropriate moment in the customer life cycle via preferred customer touchpoints”. RTIM operates primarily based on behavior.
Rise of context marketing: Brands must be prepared to make themselves known on hundreds of channels.
Prescriptive journeys build experiences over time across channels and are based on previous interactions.
While predictive journey capabilities were once initiated mostly in messaging (primarily the email channel), capabilities have expanded to include being able to automatically initiate a journey from a call center interaction or start a journey based on one’s segment membership in a DMP.
Marketers ultimately need the capability to build predictive journeys that happen over longer periods of time (journey building) and the ability to manage interactions in real time across channels (RTIM).
Marketing organizations must ensure that they have the appropriate analytical resources available to support all relevant analytical capabilities. Some of the most critical categories of support include: segmentation customer lifetime value campaign and channel measurement media mix and attribution predictive and propensity modeling social analytics dashboarding and visualization.
A sound data integration and metadata strategy is required to ensure accurate, reliable and consistent analysis, and derived insights.
The discipline of “multitouch attribution” or MTA is part science and part witchcraft but, once agreed on as the source of truth for marketing success, very hard to displace as a working model.
MTA models take several different approaches.
- A “linear” attribution model assigns equal value to every touchpoint along the way.
- “U- shaped” models favor the first and last touches more than the rest.
- “Time decay” models value each touchpoint that is closer to the sale with more value.
Media mix modeling, or MMM, can be thought of as MTA’s aggregated cousin. Media mix modeling often looks beyond the immediate time period, considering several years’ worth of media spending.
On the analytics side, whereas dashboards could only munge and visualize data for manual analysis, AI brings scale and speed to model building, by evolving dynamic models as more data comes into the system.
Just as the CDP unifies people data, the right marketing analytics dashboard unifies cross-channel interaction data. The combination of the two will create exponential value.
Business intelligence (BI) platforms go beyond marketing analytics through combining data mining, visualization, data tools and infrastructure.
We laid out the five pillars of the ideal CDP. These were:
- data ingestion from any common source and format
- data harmonization, or the ability to format and cleanse this data
- identity management, allowing users to tie information about customers together accurately
- segmentation in a user-friendly, drag-and-drop interface, as well as other analytics
- activation, or the ability to send decisions and instructions wherever they need to go for use
Marketers are going to require their CDPs to become experts in making the aggregate-level data that Google and Facebook can give them, such as “Federated Learning of Cohorts” (FLOC) or another method available to map against one’s first-party data such that they become valuable for analytics, insights and optimization.
In the near future, we will see entire sectors fundamentally changed by AI running on clean, well-structured CDP datasets.