How to select the best CDP implementation strategy for your needs
Customer Data Platforms (CDP) tout the capability to integrate multiple sources of information about potential and existing customers to enable hyper-granular segmentation, targeting, and customer experiences to help increase revenues.
To achieve those results, they need to combine several distinct modules. Depending on your business environment, data sources, and technologies currently used in your company, several possible implementation strategies exist.
Neal Analytics, a Fractal company, has experience with all three main implementation strategies: Full SaaS, Full-Custom, and Hybrid. This article describes those strategies and proposes a set of high-level criteria to help you decide which one is the most appropriate for your business goals and technological choices.
But first, to ensure the scope of what we refer to when we discuss CDP implementation strategies, let’s identify the key modules of a typical implementation.
Typical CDP architecture
A CDP implementation usually consists of a few core modules and a couple of optional ones.
CDP solution core components
A CDP implementation will always require those core modules, whether those modules are part of an existing SaaS solution or are custom-developed leveraging off-the-shelves platforms.
Data ingestion
The type and quantity of first-party data sources a CDP will ingest dramatically varies depending on the customer’s marketing and data maturity.
Those sources will mostly come from:
A key aspect of this ingestion step will be normalizing all heterogeneous data formats to enable seamless integration in the main CDP module.
Profile unification
CDPs create and manage a unified customer profile database, a “360° customer view”, by merging duplicate and heterogeneous sources into a single consumer profile.
It is a complex step as, with limited data, it can be challenging to achieve this solely through first-party, i.e., customer-owned, data.
In this situation, enriching profiles with third-party data (see “optional” section) could help the CDP deduct that?[email protected] ?and?[email protected] ?re the same person.
Segmentation
Once those unified profiles are created, the CDP can generate micro-segments for hyper-targeted marketing or sales activities.
Those segments will use one or (most often) multiple profile elements of those unified profiles. The same consumer can be part of a single activity, e.g., a promotion, or multiple parallel, sequential, or recurring ones.
Activation
Although not part of the CDP itself, whether activating (e.g., through martech tools) a customer with a specific campaign will be easy to implement or not will depend on the CDP architecture and existing (martech) connectors.
An example of this segmentation and activation aspect would be to target the segment of customers that have, in the last seven days, visited the company website twice, interacted with the website’s chat service, and asked a question on the company’s Facebook page. This profile and time-specific micro-segment could receive an email with a time-limited discount offer right before the weekend. It could then be automated and run weekly, from segment members updates to activation, without requiring human interactions.
Additional optional CDP capabilities
Most CDPs also offer the capabilities to enrich first-party data sources with third-party data sources and to expand segmentation capabilities through machine learning/AI.
领英推荐
Dataset enrichment can happen through multiple forms:
Similarly, most CDPs will offer native and third-party support for custom ML and AI models to improve further and expand segmentation capabilities beyond the ones shipped “in box” with the CDP.
CDP integration strategies
Although the integration strategies continuum is more granular than this simplified model defines, it is helpful to consider your CDP integration strategy as a choice between three different approaches:
With the full SaaS approach, all data integration, profile unification, and customer segmentation happen within the CDP. The full custom approach will skip the use of any off-the-shelf SaaS CDP solution and will focus on building a fully custom CDP that perfectly matches the customer’s needs. The last and often preferable strategy is the hybrid one. With the hybrid approach, the data ingestion and preparation will happen on an open data platform, and the CDP will be used for what is best: profile unification.
Choosing a strategy
Which of those three strategies is the best for you? As you probably guessed already, the answer is “it depends.” At Neal Analytics, we have implemented and deployed CDPs using all three approaches and experienced the unique benefits and limitations of each of those strategies.
To help you get started with your implementation strategy decision, here are key elements to consider when you need to decide on the best approach based on your needs and situation.
Full SaaS CDP implementation benefits
This approach is interesting if three conditions are met.
With this strategy, all your customer data will be stored and accessible as unified profiles in your CDP. You can also expand data source support by using the extensibility tools that all CDPs offer.
The flip side of this approach is that if you need to store and access data in a centralized data store hosted by your CDP provider. It also means that you must find or train CDP vendor-specific developers if you need new data connectors.
Full custom CDP implementation benefits
There is little need to spend much time on this approach, as it is obvious. You will fully control your data architecture and hosting with a custom CDP. You will also be able to develop unified profiles that perfectly match your needs.
The flip side of this approach is also obvious. It will take longer, require advanced data and analytics capabilities from your development team or consultants, and application maintenance will be more complex and expensive.
Hybrid CDP implementation benefits
This approach is, for most customers, often the more effective one. But not always. It offloads the step that is often the most complex and resource intensive from the CDP, and it leverages the CDP for what it is best at: profile unification.
In a hybrid approach, the source data is ingested in an open customer-owned data platform. It is manipulated and normalized before being fed into the CDP for unification. A single customer-owned (vs. CDP SaaS-based) data store will host both the source and unified data. Also, new connectors will be easier and faster to develop because an open platform is used, such as Azure Data Lake. In addition, this source and unified data will be easily accessible by any other business applications beyond the CDP.
For instance, if different departments using different tools such as?Qualtrics ,?Survey Monkey , or?Microsoft Customer Voice run ?customer surveys. With the hybrid approach, any business app or end-user-built Power BI dashboard can directly tap into this unique data lake. Without it, users would require either sharing survey tool access (with all the security and PII issues it implies) or exchanging CSV files by email.
How to get started with your CDP implementation strategy selection?
Although the options to implement a CDP are more granular than the three mentioned above, this taxonomy gives a good idea of the broad possible implementation strategies.
After you assess your short and long-term needs, internal capabilities, and available budget, you will need to map those to criteria such as data sources, residency, access beyond the CDP, extensibility requirement through ML and AI, developer availability, and time-to-value requirements. Once this need and mapping assessment are complete, you can choose the best strategy for your specific business needs and technology stage.
Neal Analytics can help you through this assessment phase and through your design, implementation, expansion, and maintenance phases. We offer flexible engagement models from assessment workshops, fixed cost or time and material projects, managed capacity, in-house team extension, and managed services models.
Learn more:
(This article was originally published on Neal Analytics blog )