Deliver Personalized CX with Unbundle CDP- Part 1
Today customers interact with a brand across a wide array of devices and expect a personalized customer experience across every touchpoint. Every touchpoint is part of an end-to-end customer journey that must provide a positive and seamless experience that brings customer satisfaction. Customer satisfaction is the main goal for an organization that wants to increase revenue, thrive and stand out in the daily market battle. The current state of personalization with the below surveys:
From the surveys, we noticed that customers often receive poor personalization and irrelevant messages with low chances to experience an end-to-end customer journey. To deliver personalized customer experience, many organizations are investing in the Customer Data Platform (CDP) that unifies customers' data from multiple channels and enables companies to deliver personalized and contextualized messages to the customers. A successful personalized customer experience is the direct link to the following four points that can be achieved by CDP:
CDP structure and state of CDP implementation
These four points introduced a typical CDP structure with four main layers data collection, data transformation, data store and modelling, and data activation.
CDPs have been around in the market with a continuous evolution that brings a huge variety of vendors. Now, how far is the evolution? Did these organizations which implemented CDP achieve their goals? Which perception did CDP build across different teams from data engineers to marketers? Did CDP bring value to the organization and what was the time to value? No doubt that organizations and vendors have made real progress throughout these years, but if we look at some research despite a vast array of vendors, we notice that organizations struggle to implement a CDP. In fact, since March 2015, Signal Group survey stated that "only 6% of marketers have a single view of the customer, despite 90% reporting it is a top priority" until the last Gartner survey from January 2022 where “only 14% of organization have achieved a Unified Profile, despite 82% who say it is still a top goal”.
Yet, another two interesting CDP Institute surveys where the first one stated that just “23% of consumer marketers have completed their projects on time and schedule. Only 58% of companies with a deployed CDP say it delivers significant value”. The second survey where said that unified profile and CDP can be implemented independently, "although 19% of all companies have them both, another 19% have a unified database without a CDP, and 13% have a CDP without a Unified Profile".
With the mentioned survey related to the Off-the-shelf CDP, it led to a question on whether we can improve our way of onboarding and implementing CDP in order to reach our CDP's goals and deliver a personalized experience.
The rise of Unbundle CDP approach
The difficulty of having a unified customer database and delivering a significant time-to-value during the CDP implementation by using the traditional Off-the-shelf CDP approach brings the rise of a new approach as a strong alternative - Unbundle CDP.
Unbundle CDP helps organizations to harness their customer data with Modern Data Stack components in lego-like building blocks and leverage the power of a data cloud warehouse. By implementing a best-in-class product at each layer of the CDP (data collection, data transformation, data store and modelling, and data activation), an organization can solve problems beyond the common use cases of Off-the-shelf CDP. In addition, understanding each of these components will allow teams to make the most informed architecture decisions when implementing their own Unbundle CDP.
Why Unbundle CDP
Unbundle CDP brings significant improvement at each layer of CDP in terms of time-to-value, flexibility, easier of scaling up, easier of iterating with decoupling components, more manageable to improve performance, and better data governance. In this article, I will focus on the three main enemies of a personalized customer experience such as lack of single source of truth, lack or low usage of 1st party data, and poor data transformation process from raw data to final data ready to be used by marketers and how unbundle CDP can solve them.
a) Lack of single source of truth : most of the time organizations that want to implement a CDP with the Off-the-shelf CDP approach, have already invested money, and efforts in the data warehouse by centralizing data in one place. The onboarding of Off-the-shelf CDP in the martech stack will inevitably bring a duplication of data and no single source of truth that it is the main requirement to have a unified customer database and break the CDP promise to eliminate silos architecture
Recommended by LinkedIn
Building Unbundle CDP around the data warehouse with a data warehouse-first approach, allows an organization to have a single source of truth where different teams can rally around the same dataset, no wasting of effort to align two different systems and last but not least build trust on data across teams.
b) Lack or low usage of first party data: since successful personalization messages require first party data, these data are usually imported into the data warehouse using the ETL pipeline, which the Off-the-shelf CDP cannot import most of the time or face limitations for security reason. Unbundle CDP allows data engineers and marketers to fully leverage first-party data to deliver personalized messages.
c) Poor data transformation process: most of the time, data engineers struggle to hydrate BI or data activation tools dedicated to the marketers cause of the monolith architecture and tight data model of Off-the-shelf CDP, which cause these issues:
These issues cause the marketer to struggle and activate data, have a quick time-to-market, and get valuable insights from data since that different work on different datasets and different business definitions.
Unbundle CDP allows data engineers to have a smooth data transformation process thanks to best in class products at each layer (data collection, data transformation, and data activation) with the possibility to replace easily a component with another one if the component does meet use cases or requirements.
Reverse ETL
Unbundle CDP introduces a new concept as Reverse ETL at the data activation layer. Reverse ETL is the process of activating and sending data from a data warehouse into business applications like CRM, social media, analytics, marketing automation, etc. There are two main excellent aspects of this layer. First, a marketer can create their audiences with a no-code approach by using a customized data model and sending it over to external systems without caring about the technical aspect. The second one, the chosen modern data stack components don't own any data and the marketer can activate and send the audience as soon as the data are available in the data warehouse.
How to implement an unbundle CDP
Based on use cases, organization martech architecture, skillsets, and organization culture. An organization can adopt a different trade-off during unbundled actions. Next, I will show three possible ways to build an unbundle CDP based on a different trade-off with a data warehouse-first approach:
Key takeaways
Go-to-Market Leader @ Hightouch | MarTech, AI, Data Products
1yAwesome post Luca Lattarini! We at Hightouch can't agree more 🚀
3x Adobe Certified | Adobe CDP | Composable CDP | Analytics | Martech Stack | Professional Google Architect Certified | Level 400 Google GenAI | Google Machine Learning Engineer | Google Data Engineering
1yThanks Desmond Phua. All mentioned will be approached during my next threes articles with these main concepts: 1)Clear roles between data engineer and marketers. Data engineer will focus on how to hydrate BI or marketing tools for marketers and Marketers will only care to activate data with no technical skills 2) Different teams rally around same dataset and have same business definition 3) focus on 1st party data to send personalized messages 4) Find business patterns to get insights from them with business driven approach during data modelling 5)Based on business use cases, data engineer will model data from raw data to final data ready to be analyze with ML/AI 6) Id stitching and Id resolution has to be adapted to the business cases with bespoke approach. From my point of view there is no one recipe for everyone but be able to provide a flexible and easy to manage solution at every organisation level it is the main goal
Digital Experience Practitioner | CX | Digital Marketing | Digital Transformation | Consumer Finance | eCommerce | Design Thinking | Experience Design | Start-ups | FSI | Fintech
1yNice article. Look forward to your next 2 article that talks about architecture. One of the failure for CDP implementation is that it is taken as a tech project. CDP should not be seen as a technical project but rather what need to be address is the overall strategy. It is good to have 360 view of the customer and have a unified profile, the question then is "so what". How is the business going to make use of that data to personalise? How do you ensure feedback loop and tweak the personalisation?... How does the CDP fit into overall MarTech strategy. At the end of the day, it is not about getting a shiny new tool but how are you going to make use of the tools to drive business goals. Once a strategy is mapped out, that is where you peel the onion further to look at data sources, type of data you need, data modelling with AI/ML, etc. Just my thoughts Luca Lattarini
Great stuff Luca!