Unbundling the customer data platform

The modern data stack offers a more flexible approach to first-party data than the conventional, all-in-one CDP.
March 29, 2022

The customer data platform (CDP) is an all-in-one marketing and data platform that promises a single view of a customer by unifying customer data from a wide range of data sources. CDPs aim to serve as a database for all your customer information with a bundled activation layer to help you leverage the data for marketing automation. When fully implemented and adopted by all teams within a company, a CDP is valuable to a business’ bottom line.

A CDP consists of both a database and software to perform the following four functions:

  1. Ingestion: Collecting events on your customers from every martech systems in use
  2. Unification: Creating a single source of truth for customer records consisting of a profile and interaction history 
  3. Analysis: Producing business intelligence and decision support by analyzing for cohorts and conversion and segment customers in audiences
  4. Activation: Pushing enriched profiles and segments to marketing channels
Boston Consulting Group published an article, “With Customer Data Platforms, One-to-One Personalization Is Within Reach,” including this slide which visualizes the four functions well.

The first two functions — ingestion and unification — result in a data warehouse containing a 360 degree view of the customer that can support any number of business scenarios: 

  • Product data - actions they've taken within your application (like what items have been viewed or left abandoned in their shopping cart)  
  • Sales data - where a customer is in their buying journey
  • Marketing data - what blog posts they've read, how they first landed on your site, etc. 

The third and fourth functions — analysis and activation — turn data into action by segmenting consumers who can be activated through the appropriate channels, such as email, SMS, CRM, advertising and on page personalization.

With the ongoing disappearance of anonymous customer identifiers such as cookies and device IDs, a first-party customer data platform is more essential than ever.

Unfortunately, a bundled, all-in-one CDP is a black box that creates serious rigidities and complex workflows that are difficult for a company to use.

The trouble with CDPs

Unfortunately, not all CDPs are fully implemented and adopted by the teams they are intended to serve. The CDP Institute’s latest survey found that just “23% of consumer marketers have completed their projects on time and on schedule and only 58% of companies with a deployed CDP say it is delivering significant value.”

Even when fully implemented, CDPs introduce new and often redundant work to organizations where people and resources are already taxed with other pressing full-time responsibilities. This usually results in the three following complaints:

1. “I don’t have time to use the CDP”

Throughout my career, I heard this from marketing teams whose activation flows now required extra steps due to the new CDP. Audience segmentation is usually handled in activation channel tools — advertising, email, on-page personalization and so on. After adopting a CDP, teams have to rebuild audiences and segments using an upstream tool. CDPs can create more complicated, duplicative workflows, not to mention impose barriers to entry for a team in the guise of learning a new tool.

2. “I don’t trust the data in the CDP”

I often heard this from data and BI teams who didn't trust the CDP's profile unification model, as off-the-shelf CDPs are a black box in terms of modeling. One BI team I worked with was tasked with reporting on the total first-party audience using the deduplicated and unified set data captured within the CDP database. In our case, both the CDP and the BI team used Google BigQuery, so data syncs happened in real-time between the two databases. Unfortunately, once the BI team started working with the data, they questioned the unification model so much so that they re-ingested the same data sources like ESP to build their own model.

The black-box nature of the CDP leads analysts to distrust models and calculations created by the CDP. They can’t audit or otherwise vouch for their integrity. This produces yet more redundant work as they end up separately rebuilding the data models.

3. “I prefer my current workflow over the CDPs” 

I heard this from both data and marketing teams. In most cases, the internal teams were already performing the critical functions, albeit with different tools and technology.

  • Data Science and BI teams were already ingesting, normalizing and analyzing data designed to produce insights and reports for the business.
  • Marketing teams analyzed, segmented and activated audiences, albeit from a much smaller data set.

All-in-one CDPs often duplicate the functionality of existing tools and workflows for no real gains in performance or usability.

I experienced these issues first hand in 2019 at Universal Music Group (UMG). UMG’s digital touch-points numbered in the thousands as marketing occurred through separate artist sites, stores and email lists.

In aggregate, Universal Music Group had millions of first-party data points about their consumers, but the complex technical environment made it impossible to answer basic questions like:

  • How big is the combined and deduplicated audience?
  • How can we identify and engage with the biggest fans of our artists?

We implemented our CDP over six months, integrating CMS, ecommerce online marketing, email service provider systems and more. For all our effort, neither the reporting nor activation teams adopted and utilized these new capabilities after they launched.

The solution: An unbundled approach to CDP

A more flexible and modular approach to customer data management relies on "unbundling" the functions of the CDP. In effect, all the elements making up a SaaS CDP can now be decoupled from a single vendor and separately implemented by the appropriate sections within a business. It means less work, less redundancy and more adoption by the company at large. 

The four functions performed by a CDP correspond to a number of existing tools that make up the modern data stack:

  1. Ingestion – Data integration and data pipeline tools like Fivetran
  2. Unifying the Customer Record – Data transformation tools like Fivetran data models (built using dbt Core by dbt Labs)
  3. Intelligence and Decision Making – Business intelligence tools such as Looker or Tableau
  4. Activation – Reverse ETL like Hightouch

Iterative change

Decoupling the functions of a CDP frees a company up to iteratively test digital tools rather than attempt a significant digital transformation process in one fell swoop.

For example, automated data pipelines usually feature connectors for common data sources, such as SaaS apps and can easily handle the ingestion function of the CDP. Many data pipelines include or are readily compatible with transformation tools, which can unify the customer record. Common business intelligence platforms can aid decision making by massaging data models into easily digestible visualizations and dashboards. Finally, reverse ETL can easily sync customer data back to marketing tools activation. 

Unbundled tools mean marketers can continue their existing workflows but with an enriched data set. Contrast this approach with the heavy lift of deploying a CDP from scratch, where all systems need to be integrated into a new database (typically 3-6 months minimum) before the marketing team can start activating data. 

An unbundled approach requires careful management of a data warehouse, but a data warehouse should be the central part of any organization’s data stack anyway.

Using the MDS to build your CDP

While I've focused on some of the challenges faced in CDP deployments, I would never say it is not worth the effort. A first-party data solution is essential for every company that sells directly to customers. It was important five years ago and has only increased in importance with the post-pandemic growth of e-commerce and the demise of cookie and device identifiers.

Some companies may continue to use a packaged CDP product but savvier ones will use the modern data stack to leverage their investments to power CDP functionality like personalization, growth and performance marketing from their existing teams and technology. In short, the CDP landscape is changing, but all for the better.

Want advice on the best way to build your own CDP? Download The Data Professional's Guide to Data Integration to learn more. 

Commencer gratuitement

Rejoignez les milliers d’entreprises qui utilisent Fivetran pour centraliser et transformer leur data.

Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
Data insights
Data insights

Unbundling the customer data platform

Unbundling the customer data platform

March 29, 2022
March 29, 2022
Unbundling the customer data platform
SUJETS
No items found.
PARTAGER
The modern data stack offers a more flexible approach to first-party data than the conventional, all-in-one CDP.

The customer data platform (CDP) is an all-in-one marketing and data platform that promises a single view of a customer by unifying customer data from a wide range of data sources. CDPs aim to serve as a database for all your customer information with a bundled activation layer to help you leverage the data for marketing automation. When fully implemented and adopted by all teams within a company, a CDP is valuable to a business’ bottom line.

A CDP consists of both a database and software to perform the following four functions:

  1. Ingestion: Collecting events on your customers from every martech systems in use
  2. Unification: Creating a single source of truth for customer records consisting of a profile and interaction history 
  3. Analysis: Producing business intelligence and decision support by analyzing for cohorts and conversion and segment customers in audiences
  4. Activation: Pushing enriched profiles and segments to marketing channels
Boston Consulting Group published an article, “With Customer Data Platforms, One-to-One Personalization Is Within Reach,” including this slide which visualizes the four functions well.

The first two functions — ingestion and unification — result in a data warehouse containing a 360 degree view of the customer that can support any number of business scenarios: 

  • Product data - actions they've taken within your application (like what items have been viewed or left abandoned in their shopping cart)  
  • Sales data - where a customer is in their buying journey
  • Marketing data - what blog posts they've read, how they first landed on your site, etc. 

The third and fourth functions — analysis and activation — turn data into action by segmenting consumers who can be activated through the appropriate channels, such as email, SMS, CRM, advertising and on page personalization.

With the ongoing disappearance of anonymous customer identifiers such as cookies and device IDs, a first-party customer data platform is more essential than ever.

Unfortunately, a bundled, all-in-one CDP is a black box that creates serious rigidities and complex workflows that are difficult for a company to use.

The trouble with CDPs

Unfortunately, not all CDPs are fully implemented and adopted by the teams they are intended to serve. The CDP Institute’s latest survey found that just “23% of consumer marketers have completed their projects on time and on schedule and only 58% of companies with a deployed CDP say it is delivering significant value.”

Even when fully implemented, CDPs introduce new and often redundant work to organizations where people and resources are already taxed with other pressing full-time responsibilities. This usually results in the three following complaints:

1. “I don’t have time to use the CDP”

Throughout my career, I heard this from marketing teams whose activation flows now required extra steps due to the new CDP. Audience segmentation is usually handled in activation channel tools — advertising, email, on-page personalization and so on. After adopting a CDP, teams have to rebuild audiences and segments using an upstream tool. CDPs can create more complicated, duplicative workflows, not to mention impose barriers to entry for a team in the guise of learning a new tool.

2. “I don’t trust the data in the CDP”

I often heard this from data and BI teams who didn't trust the CDP's profile unification model, as off-the-shelf CDPs are a black box in terms of modeling. One BI team I worked with was tasked with reporting on the total first-party audience using the deduplicated and unified set data captured within the CDP database. In our case, both the CDP and the BI team used Google BigQuery, so data syncs happened in real-time between the two databases. Unfortunately, once the BI team started working with the data, they questioned the unification model so much so that they re-ingested the same data sources like ESP to build their own model.

The black-box nature of the CDP leads analysts to distrust models and calculations created by the CDP. They can’t audit or otherwise vouch for their integrity. This produces yet more redundant work as they end up separately rebuilding the data models.

3. “I prefer my current workflow over the CDPs” 

I heard this from both data and marketing teams. In most cases, the internal teams were already performing the critical functions, albeit with different tools and technology.

  • Data Science and BI teams were already ingesting, normalizing and analyzing data designed to produce insights and reports for the business.
  • Marketing teams analyzed, segmented and activated audiences, albeit from a much smaller data set.

All-in-one CDPs often duplicate the functionality of existing tools and workflows for no real gains in performance or usability.

I experienced these issues first hand in 2019 at Universal Music Group (UMG). UMG’s digital touch-points numbered in the thousands as marketing occurred through separate artist sites, stores and email lists.

In aggregate, Universal Music Group had millions of first-party data points about their consumers, but the complex technical environment made it impossible to answer basic questions like:

  • How big is the combined and deduplicated audience?
  • How can we identify and engage with the biggest fans of our artists?

We implemented our CDP over six months, integrating CMS, ecommerce online marketing, email service provider systems and more. For all our effort, neither the reporting nor activation teams adopted and utilized these new capabilities after they launched.

The solution: An unbundled approach to CDP

A more flexible and modular approach to customer data management relies on "unbundling" the functions of the CDP. In effect, all the elements making up a SaaS CDP can now be decoupled from a single vendor and separately implemented by the appropriate sections within a business. It means less work, less redundancy and more adoption by the company at large. 

The four functions performed by a CDP correspond to a number of existing tools that make up the modern data stack:

  1. Ingestion – Data integration and data pipeline tools like Fivetran
  2. Unifying the Customer Record – Data transformation tools like Fivetran data models (built using dbt Core by dbt Labs)
  3. Intelligence and Decision Making – Business intelligence tools such as Looker or Tableau
  4. Activation – Reverse ETL like Hightouch

Iterative change

Decoupling the functions of a CDP frees a company up to iteratively test digital tools rather than attempt a significant digital transformation process in one fell swoop.

For example, automated data pipelines usually feature connectors for common data sources, such as SaaS apps and can easily handle the ingestion function of the CDP. Many data pipelines include or are readily compatible with transformation tools, which can unify the customer record. Common business intelligence platforms can aid decision making by massaging data models into easily digestible visualizations and dashboards. Finally, reverse ETL can easily sync customer data back to marketing tools activation. 

Unbundled tools mean marketers can continue their existing workflows but with an enriched data set. Contrast this approach with the heavy lift of deploying a CDP from scratch, where all systems need to be integrated into a new database (typically 3-6 months minimum) before the marketing team can start activating data. 

An unbundled approach requires careful management of a data warehouse, but a data warehouse should be the central part of any organization’s data stack anyway.

Using the MDS to build your CDP

While I've focused on some of the challenges faced in CDP deployments, I would never say it is not worth the effort. A first-party data solution is essential for every company that sells directly to customers. It was important five years ago and has only increased in importance with the post-pandemic growth of e-commerce and the demise of cookie and device identifiers.

Some companies may continue to use a packaged CDP product but savvier ones will use the modern data stack to leverage their investments to power CDP functionality like personalization, growth and performance marketing from their existing teams and technology. In short, the CDP landscape is changing, but all for the better.

Want advice on the best way to build your own CDP? Download The Data Professional's Guide to Data Integration to learn more. 

Topics
No items found.
Share

Articles associés

No items found.
No items found.
No items found.

Commencer gratuitement

Rejoignez les milliers d’entreprises qui utilisent Fivetran pour centraliser et transformer leur data.

Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.