Customer Data Platforms
Data & AnalyticsMarketing

Customer Data Platforms

Martin Kihn, Christopher B. O'Hara
Read 7 August 2021

Review

I read this while my team was evaluating martech infrastructure. CDPs kept coming up in vendor conversations and internal debates, and I did not have a clean mental model for where they fit. This book provided one.

Kihn and O'Hara map the full landscape: how a CDP relates to a CRM, a DMP, and a data warehouse; what the five core capabilities are and how they build on each other; and why most organizations have fragments of this working but not the whole thing. The structural clarity is the main contribution. It is more reference text than practitioner playbook, but as a shared vocabulary for a complex category, it holds up.

The consent and GDPR sections are worth reading separately from the rest. The seven GDPR principles are presented with practical intent rather than legal defensiveness, and the framework for what earns consumer trust sits alongside the compliance requirements in a way that makes the whole thing feel actionable.

Some chapters feel like extended vendor pitch material. The data science chapter covers supervised versus unsupervised learning at a level that would not challenge anyone who has worked with data. But the core argument, that data unification is the real infrastructure problem and that CDPs exist to solve it, is argued well and backed with enough specifics to be useful.

Key Takeaways

The parts worth keeping:

Why data silos exist and what they cost

CRM and loyaltySalesforce, AWS, SQL Server
Email, mobile, socialSalesforce Marketing Cloud, LinkedIn, Instagram, YouTube, Facebook
Sales and commerceSalesforce Commerce Cloud, Magento, SAP HANA
Web analyticsAdobe Analytics, Google Analytics
Paid digital adsGoogle Ads, Amazon, Instagram, Snapchat
Programmatic adsThe Trade Desk, AppNexus, TubeMogul
Customer databasesData lakes, propensity scores, internal models
Agency reportingExcel, Google Sheets, PDFs

How CDP fits alongside other tools

Data WarehouseCustom IntegrationDMPCRMCDP
Unified customer dataYesNoNoPartialYes
Persistent storageYesYesNoYesYes
Packaged systemNoNoYesYesYes
Real-time capabilityNoPartialYesNoYes
Open accessYesNoYesPartialYes

The CDP is an evolution of the CRM category, not a replacement. It operates as the platform that marketing and advertising systems plug into, with customer data as the connective layer. A platform, not a product.

The five CDP capabilities

CapabilityWhat it does
Unified ProfileThe destination for everything else. Assembles fragmented records from CRM, loyalty systems, call center, and third-party sources into a single customer record. Fidelity matters as much as coverage: an accurate but incomplete profile is more useful than a comprehensive but stale one.
Data CollectionIngests data from multiple sources through batch, streaming, and event-based pipelines. Covers both known identifiers (PII) and unknown ones (cookies, device IDs) from the first customer interaction. The collection layer exists to serve the profile.
Data ManagementHarmonizes what the collection layer brings in. Different systems name the same field differently, store dates in different formats, and carry duplicate records. This layer maps, cleanses, and deduplicates so the profile can be trusted.
Segmentation and ActivationTurns the unified profile into targeted audiences and distributes them to email platforms, paid media, call centers, and commerce systems. Centralizing segmentation means the same audience definition is available across every channel simultaneously.
Insights and AIThe intelligence layer on top of the profile. Predictive scoring, ML-built segments, next-best-offer recommendations, and engagement timing models. The CDP is the natural source for any model that needs a complete customer view.

Identity resolution: known versus unknown

The seven GDPR principles

  1. Transparency: Customers must know what data is being collected and by whom, before collection happens.
  2. Purpose limitation: Data collected for one stated purpose cannot be repurposed for something the customer did not agree to.
  3. Data minimization: Collect only what the stated purpose requires. Excess data is a liability, not an asset.
  4. Accuracy: Data must be kept current. Stale records that misrepresent the customer undermine both compliance and marketing effectiveness.
  5. Storage limits: Retaining data indefinitely is not compliant. Retention policies need end dates tied to the original purpose.
  6. Security: Collected data must be protected from unauthorized access. Both a legal obligation and a basic trust requirement.
  7. Accountability: The data controller must demonstrate compliance, handle subject access requests, and allow customers to correct or delete their records.

Organizational prerequisites