Europe/U.S. Roundtable, April 23, 2025

Guest: Steve Zisk, Product Marketing Principal, Redpoint Global

Topic: Data Readiness– Steve Zisk

CDP issues

• Declining customer experience satisfaction (peaked in 2020 per Forrester)
• Low utilization of martech systems (more than 60% of companies have CDP in place but less than 15% are seeing expected value from their CDP per Gartner)
• Data is not AI-ready (per Gartner)

One reason for issues is that CDP isn’t providing enterprise-ready customer data

• Data needs to be right (complete, accurate, timely) and fit for purpose (actionable, trusted, compliant)
• Users often don’t understand what they need to do to get their data ready; are trying to educate clients and prospects what data readiness is and why it’s important for CDP and for all kinds of martech and other customer-focused technology and projects

CDP Data issues

• Data silos: part of reason is confusion among buyers of what a CDP does
• Timeliness: data may not be timely enough for given use case
• Fragmented profiles, due to failure to link all customer IDs

Causes of CDP dissatisfaction

• People may be unhappy with their CDP because they don’t understand what it really is, or think they have one when they don’t
• People solve one initial use case with CDP but don’t continue to develop new use cases
• Full utilization requires people and processes as well as technology; this applies to any martech
• Clients sometimes set up ingestion incorrectly initially, because they don’t understand nature of their data;
• Clients may not realize source data must be suitable for CDP e.g.
  • values must be consistent (titles, currency, measurement systems, data entry errors, incomplete entry),
  • don’t correctly unify profiles when people use different communication channels/devices (e.g. different email accounts);
  • same account may be used by multiple people, may share credit card, ship to different addresses e.g. child at college; system needs to allow complex relationships (at time that set up system, or have to redesign later)
  • composable CDP distributes a problem across multiple modules (e.g. assume that data quality is solved before data enters warehouse)
• fairly common for company to need to redesign their CDP database after initial implementation
  • clients often don’t have access to source data in advance of deployment
  • data changes over time, need to keep up with that (‘observability’ is IT term for monitoring data for availability, quality, etc.)
  • may need to do redesign to gain additional value beyond initial use cases
  • users may need to redesign because they did not originally understand what they needed (e.g. correct householding rules)
  • is often initial rush to deploy, which leads to mistakes if users do not correctly understand their requirements; especially for identity resolution
  • extends beyond customer data, e.g. also product data, real time data
  • is significant learning curve for every aspect of CDP and activation
  • is need for continuous improvement, in addition to initial training; users must consider next steps, how to expand user community, number of user cases, etc.
  • vendors should set expectation for continuous improvement, but RFPs often ask for quick initial results, which deters vendors from explaining need for long-term development 
• clients recognize need for AI-ready data; is opportunity for CDP vendors to solve problem
  • Gartner expects more than 30% of generative AI projects to be abandoned due to poor data quality, inadequate risk controls, escalating costs, or unclear business value
  • 40% of users cite “lack of data” as top-3 barrier to implementing AI; is really about lack of accessible data
  • Distinguish gen AI data for creating content (e.g. to train generative AI) vs data to activate AI (e.g. personalization); need both
  • AI needs more than just customer data e.g. product information, location, site, persona, user goals (e.g. research vs purchase vs service)
• Agentic AI will possibly solve some problems e.g. selecting and finding the right data, but could create new problems e.g. identify data types (need semantic content for AI to understand what is in the data)
  • AI agents are likely to be specialists that need to work with other agents e.g. MCP (model context protocol)
  • CDP could become MCP server; CDP could provide other services to agents e.g. identity resolution, data cleaning, address verification; is a place where composability makes sense
  • MCP and agentic are immature; are still many practical issues e.g. security, guardrails, etc. before they deliver expected value
  • Agents will be able to help with analytical tasks such as analyzing a/b tests, finding what top customers are buying, drilling into data, building segments, etc.
  • AI needs to explain ‘why’ something happens, and to explain its own thought process for why it reached a conclusion
• Data quality, access, etc. is role of data team and warehouse,
  • Data team may see CDP and other martech as ‘toy’, not ‘first class’ citizen; may think AI can replace CDP at CDP specialist tasks such as identity resolution
  • Original conception of CDP was to provide self-service tool for marketers to interact with customer data;
  • as value of data has been recognized, data has become more of an enterprise asset and been more privacy and compliance requirements
  • this means customer data needs to be more carefully governed,
  • can’t only be managed by marketers, but marketers still need self-service access
  • been increased involvement of IT, finance, data teams in CDP buying decisions

APAC Roundtable, April 23, 2025

Guest: Hari Prasath, Senior Manager, Support – APJ, Tealium

Topic: Factors for Successful CDP Utilization

CDP architecture has three major components: data sources, CDP itself, which assembles customer profiles, and activation

• Within CDP, are components for ingestion (client and server-side), profile attributes (content browsing, engagement/intent, consent preferences, account data, user/device ID, 2^nd/3^rd party data), audience segmentation (product based, engagement based, account-based), AI/ML models (likelihood to convert, to return), transformations, insights and reporting, and profile access (real-time API)

Utilization comes at the adoption stage of the lifecycle, after the CDP is activated and handed over to team who will maintain and use it

• Three pillars of adoption: platform, process, and people
• Team design depends on team size & expertise, martech stack sophistication & existing integrations, who owns customer data, who owns business goal / budget
  • Avoid ‘land grab’ mentality and focus on building ROI
  • Common team setups:
    • IT led (59% of companies, per Tealium 2025 State of the CDP study): where multiple departments use CDP; common among gaming/betting companiesMarketing-led (22%): where marketing own martech stack: common in retail and educationData/analytics-led (18%: when data is primary focus; common when customer data systems are managed by data team
    • CDP Center of Excellence (1%): dedicated team to manage all functions of CDP; common among enterprise multi-national corporations, multiple product lines
  • Marketing-led CDPs rose from 12% in 2024 survey to 22% in 2025 survey
  • Marketing team owning activation layer increases CDP utilization, helps IT to understand needs
  • IT-led may result in marketing not defining use cases clearly, which leads IT to struggle
  • Team member roles: team led, business analyst (translates between marketing and IT), engineer (configure and manage)
• Data collection: often rushed during implementation, not monitored for subsequent changes in sources.  To avoid, must
  • define ownership of collection process.
  • Check that connectors remain current when source changes
  • Provide schema-level collection detail during RFP process (beyond simply listing data sources)
  • Continuously review compliance, since requirements may change over time
  • Server-side collection becoming more important, especially CAPI (Conversion API, used to collect responses when browser-based collection is blocked; still requires consent)
• Cross-team communication / evangelism. 
  • Define use cases to drive utilization
  • Roadshows and workshops to show users what data is available in CDP
  • Set up notifications e.g. in Slack, about events such as file imports
  • Encourage user certification (from vendor) to drive utilization by ensuring they know how to perform specific tasks; are separate certifications for end-users vs engineers
• Platform – continuous improvement
  • Regular architectural audits and reviews, to check for broken integrations, current data sources, schema updates, unused components, etc.
  • Review against maturity frameworks for gaps, relative maturity, to help plan future development
  • Automated monitors and alerts (e.g. via Slack notifications)