How to Select the Right Customer Data Platform for your Business

As organizations look to unify, segment and activate customer profiles across channels and business units, they are faced with sorting through the many customer data platforms (CDPs) on the market that fit their unique needs. CDP options vary from building a custom platform in house (for example, using Databricks or Snowflake on the enterprise’s preferred cloud) to leveraging one of many options offered by Salesforce, Adobe, Sitecore, etc. There are five major considerations that your data and analytics teams should examine when deciding the strategy and platform on which to build their CDP:

1. Activation Timelines

The question to ask here is the level of urgency to activate CDP curated data sets and profiles to enable marketing campaigns, use cases and other business objectives. The challenge with this is that response from most business units is usually “yesterday”, “next quarter” or “in six months”.  That’s why it’s imperative to take a step back and assess whether business use cases driving the CDP program are already well-defined with metrics, segments, tactics and approaches. This exercise will inform the level of business readiness and enable teams to identify whether to lean towards a custom in-house solution (full flexibility) or a vendor product (faster time-to-market).

2. Integration Requirements

After starting with the activation targets underpinning the business use cases, it is important to identify as many upstream and downstream integrations that will be required to realize segments and flow of data from and to the CDP. Once these integration points are identified, use the below to weigh your CDP options:

• Connectors: What is the availability and ease of using out-of-the-box connectors? While most modern platforms provide a plethora of connectors, it is important to check if there is an integration point that is very organization or industry specific. For example, Veeva CRM is widely used in the life sciences industry and having a connector that automatically syncs your CDP with it would make DataOps less manual.
• Pre-processing: Do you need any pre-processing or hosting areas for cleaning and prepping the shape of data before ingesting into the CDP? This is likely required for legacy systems that might only be able to export data to limited destinations, such as file transfer protocol, in certain formats or smaller sizes.
• Dependencies: Do any of your integration sources or destinations have their own integrations that need to be either ingested or activated to? For example, an eCommerce website might use Stripe for payments processing and there might be a need to integrate with their Stripe platform for payments metadata. 

3. Unified Customer Profile

In addition to data and platform integrations, the other big piece that needs to be assessed, designed and productized is the curation of unified customer profiles. Of the many data products and data assets that can be curated through the CDP, there should be no debate that the unified customer profiles is the most significant data product. While most teams understand at a high level how profiles are unified, they are certain nuances that can help you choose a CDP that works best for your requirements and use cases:

• Merge Rules: The important question to ask here is how complex the logic and rules need to be to merge customer profiles into a unified version. Will the standard out-of-the-box configuration options provided by vendor CDPs suffice? Or does custom logic need to be implemented to merge profiles? Custom logic might be needed for scenarios where an organization has made, or plans to make, a major acquisition and the acquisition’s customer data might have its own enterprise customer ids that now need to be mapped and merged with the acquirer’s enterprise customer ids.
• Refresh Frequency: How often do the unified customer profiles (and the associated segments) need to be refreshed? This can be tricky to implement both in custom CDPs (high compute costs), as well as vendor CDPs which usually have their own SLAs and work best with their stack of products. 
• Updates & Corrections: Typically, the curation of unified customer profiles is the biggest challenge. While selecting your CDP of choice, it is important to weigh how easy or difficult it is to fix bugs, make corrections and updates to profiles already merged. Ensuring that the platform supports multiple environments, GitHub integrations, ability to run CI/CD pipelines and options to roll back scripts/configurations to previous versions is a big plus. 

4. Team Skills & Maturity

Regardless of which option you choose, every CDP team will require a combination of skills in roles including data engineers, DataOps analysts, data stewards, project managers, data analysts, artificial intelligence (AI)/machine learning (ML) developers, business intelligence (BI) experts, etc. Current team skills, maturity level and hiring plans should also be considered while weighing and selecting a CDP. Having preexisting expertise in the platform, through other projects and investments, can make it easier for the team to ramp up quickly on the selected CDP and show value earlier in the implementation cycle. One caveat here when it comes to assessing a custom in-house solution versus a vendor solution is the assumption that a custom CDP team would need more data engineers and data quality experts than a team that implements and configures a vendor CDP. If the vendor option requires additional configurations or extensions (for example, for defining custom merge rules) or if there is a limited supply of professionals with exposure to that specific vendor’s CDP, then a vendor CDP option may also require more headcount than anticipated.

5. Cost

There are multiple factors that need to be considered (in addition to typical licensing costs) to truly assess and compare CDPs based on whether the cost justifies the estimated business value and maximizes ROI:

• Storage: Consider this across all various tiers (bronze, silver, gold) of storage that may be needed, as well as data retention and archiving needs.
• Compute: Since compute costs can quickly add up, especially for custom CDPs, it is best to estimate these by adding additional buffers for unplanned usage to avoid surprises. While we haven’t discussed AI/ML implementations in detail that need to be run on top of unified customer profiles to create segments and identify clusters, don’t forget to consider additional compute costs for any AI/ML scenarios that you may have already prioritized.
• Profiles: Most vendor CDPs base their cost on the number of unified customer profiles that will exist within the CDP. It is important to understand that not all profiles get merged into a unified profile (due to data quality, cookie ids, etc.) so this count will be higher in practice than the total customer base you may have.
• API Usage Volumes: Depending on the CDP, it is important to account for expected API usage and associated volumes. Even though it can be difficult to estimate these upfront, using a proxy to estimate a range can help avoid surprises later.
• Learning & Development (L&D): In addition to the actual cost of CDP, don’t forget to include costs that enable the team to learn and get certified on the platform. Having this cost included upfront makes it seamless and motivating for team members to ramp up on the platform without going through tedious approval processes. Some vendors might be open to providing discounts on their L&D packages during initial contract negotiation and it is best to capitalize on it during this stage.
• Consulting & System Integrator Costs: If you think you might benefit from engaging a consulting and systems integrations partner to either accelerate the implementation or align the design with industry best practices, it is best to also estimate these costs upfront to ensure a smoother execution roadmap. 

While these considerations are certainly not exhaustive, they provide a good starting point for companies trying to sort through the many CDP offerings currently on the market. In doing so, enterprises will be better prepared to make the right choice for their specific business needs.