When it comes to AI, a CDP alone isn’t enough
News, AI

When it comes to AI, a CDP alone isn’t enough

Chiara McPhee
July 8, 2024
Reading Time: 4 minutes

CDPs were supposed to be the answer to all our customer data woes. Don’t get us wrong; they were a huge step forward for marketers by enabling the structuring and analysis of scattered customer data. However, with advancements in generative AI, CDP data alone is no longer sufficient to power a best-in-class marketing program.

Customer Data Platforms and the core problem they were built to solve

Let's start with some history. Customer Data Platforms (CDPs) were designed to address the problem of fragmented customer data. Brands often had data dispersed across various tools—support tools, sales tools, marketing tools, and more—and sought a 360-degree view of each customer to which they marketed. Adopting a CDP meant gaining access to this comprehensive view, along with cleaner, structured data that could be used to send more targeted marketing messages.

So how did these early CDPs work? They consisted of ETL (Extract, Transform, Load) connections that allowed brands to integrate data from different systems into a central data warehouse or repository, where they could keep customer profiles up to date. You still with me? 

So Much Promise, So Little Revenue Impact

Predictive AI is machine learning trained on structured data, such as customer purchase histories, demographic information, and behavioral data collected through various channels (and stored in CDPs!). Predictive AI can make future predictions based on historical data, such as identifying high-importance customers, predicting customer churn, or recommending products.

Predictive marketing capabilities enhance the value of CDPs by enabling marketers to target customers more precisely. When predictive AI became available many years ago, marketing technologists and the marketers they served believed they had found the promised land—technology to target customers at the perfect time with the perfect product (believe me, I was one of them). Unfortunately, these models typically resulted in single-digit increases in conversion.

Unstructured Brand Data? No Longer out of Reach

This raises a crucial question: how can we better use data to skyrocket marketing conversions? Marketers have a goldmine of customer data at their fingertips, but they’ve missed one vital piece—brand context relevant to each customer.

Let me break it down. Crafting the perfect message requires two things: a) understanding the customer’s historical relationship with your business through structured data from a CDP, and b) weaving in brand context that resonates with the customer, the products you sell, and your brand voice.

Your branded experiences are your secret sauce—they’re what drive customer loyalty, recognition, and market differentiation. They’re often the reason shoppers choose to buy via Shopify-powered sites over Amazon. For a message to truly resonate, it must be relevant. And to achieve that relevance, you need a real-time grasp of the customer, the product, and the brand context sourced from the vast expanse of the internet and social media. 

Here's the kicker: this is unstructured data that, until recently, was beyond our reach. Unstructured brand context data—like social media posts, customer reviews, and news articles—paint a complete picture of your brand's online perception. This data is critical to crafting marketing messages that not only resonate but also convert.

We’re Closing In: Enter Large Language Models (LLMs)

LLMs (Large Language Models) are powerful AI systems that excel at processing and understanding unstructured data, including social media posts, current events, and various merchant and brand data scattered across the internet. These models can generate human-like text and provide concise summaries. Initially, LLMs found their niche in customer support, where they listened to conversations and suggested responses. Recently, companies like OpenAI have advanced LLM technology, enabling these models to efficiently process vast amounts of internet data and expand their applications significantly (i.e., Chat GPT).

Marketers have traditionally managed brand context and crafted brand voice manually across all marketing channels. However, the advent of LLMs and Generative AI has revolutionized what’s possible. When combined with predictive technology, these technologies can craft highly relevant messages that align with your brand's unique voice. By training AI on structured customer data (purchase history, message engagement history) and unstructured brand context data (social posts, customer reviews, news articles), marketing platforms can help marketers create content that is on-brand and contextually resonates with customers.

Predictive AI + Generative AI = Everything Marketers Have Been Waiting For

While both LLMs and predictive machine learning aim to enhance customer interactions and optimize marketing efforts, LLMs provide advanced capabilities in understanding and generating human-like text from unstructured data, whereas predictive machine learning on CDPs focuses on making data-driven predictions and insights from structured data.

Structured data in CDPs is valuable, but to achieve truly meaningful revenue lift (and by this, we mean double digits), marketers need a platform that integrates both customer and brand context data in its AI capabilities. The future of 1:1 personalization in marketing is at the intersection of brand data feeding generative AI and customer data feeding predictive AI. At Postscript, we’re developing a powerful way to do just that. Register here to join us for the big reveal.


  • CDPs' Limitations: Customer Data Platforms (CDPs) are great for structuring scattered customer data but are no longer sufficient alone due to advancements in generative AI.

  • The Role of Predictive AI: Predictive AI enhances CDPs by using structured data to make future predictions, like identifying high-importance customers and predicting churn, but often results in only modest conversion increases.

  • Importance of Unstructured Data: Effective marketing requires integrating unstructured data, such as social media posts, customer reviews, and news articles, to provide a full picture of brand context and drive relevance in messaging.

  • Power of Large Language Models (LLMs): LLMs excel at processing unstructured data and generating human-like text, enabling marketers to craft messages that resonate with customers and align with the brand's voice.

  • Future of Marketing: Combining predictive AI (structured data) with generative AI (unstructured brand context data) is key to achieving significant revenue lifts and true 1:1 personalization in marketing. Postscript is developing a platform to harness this potential. Register for our upcoming webinar to learn more.

Chiara McPhee
Chiara McPhee
Chief Product Officer