Blog

Notes from CDP Institute October Roundtables

October 28, 2024

The CDP Institute organizes monthly online roundtable discussions for all community members.  Here are notes from our October sessions.  Keep an eye out for email and calendar announcements of the next set of roundtables.

Europe/Americas Roundtable October 23

Guests:

– Roy Wollen, President, Hansa Marketing Services and Lecturer, Northwestern University

– Brian Gruidl, VP Email Marketing, Hansa Marketing Services

What is personalization

– personalization is not about knowing name, but how to shape experience e.g. show relevant items; just showing name is disappointment because it promises personalization

– 71% of customers frustrated when shopping is not personalized

– personalization can be physical, digital or both (“converged”).  Mobile and augmented reality in particular offer opportunity for new experiences.

– Best personalization is subtle, relevant and timely.  Example: when marketing manager questioned value of personalization, found ones she liked were based on her behaviors.

– want personalization to create experience that shows not just product or category but also interesting things along the way, similar to walking through store to find milk placed in the back.  CDP provides a lot of data to do this. 

– email personalization examples:

  • Spotify uses past listening habits and location, and shows when favorite artist is coming to local event
  • email from vet that includes picture of own cat
  • energy company showing how energy usage compared with others, and pattern by month, promotes energy audit; builds long term engagement shows energy company as resource

– non-email personalization examples:

  • company selling bus coach tours, uses CDP to offer tours that are near where they live, had big increase in uptake.
  • augmented reality to try on eyeglasses
  • reminders from store loyalty program when pass by location.

– Hansa has good/better/best approach:

  • good is showing best sellers to someone in a category they had shopped;
  • better is recommending specific content based on customer actions and product history;
  • best is using AI driven personalization based on huge amounts of data to create microsegments and to react to shifts in data with adjusted messaging on an individual basis.

Personalization planning

– build portfolio of use cases, then review with all stakeholders in organization to build consensus on which are important.  Need to do ongoing measurements of what actually happened vs expectations, as way to define success.

– important to consider advanced as well as simple use cases when defining requirements, so select system that can meet needs beyond the original (simple) use cases.

– create a roadmap based on what’s currently available, what would like to have, how to close the gap, and what’s the cost and value of doing that.

– don’t have to completely tear apart existing approaches.  Start with what are doing today, where you want to be, what data you have and what’s missing which could be valuable.   AI and CDP are starting to be able to identify which data elements are valuable and how much is missing, so can build plan to fill gaps and define use cases, which may involve new technology.  Can look for quick wins without tearing everything up.

– when evaluating CDP, don’t start with capabilities.  Rather, start with needs analysis for where you are, what you need, where your company is in a brand and customer perspective, and then think about what needs you have.  Do stakeholder interviews to build consensus, then look at CDP capabilities to see which are important based on your needs.  Gives a framework for evaluating CDP capabilities. 

– example: did use case analysis, identified important journeys and pain points within journeys, what is missing to enable journeys.  Realized that current technology does not meet real time needs, which gave purpose-driven use cases that could be assigned an estimated value.  Discovered would need three CDPs to meet all requirements. 

Personalization execution

– personalization requires understanding your customers, which means using your data to develop your personalization strategy and materials.

– at Target, Brian had internally-built CMS with lots of content tagged by category, product, promotion, etc.; then scored audiences against those daily; was huge processing effort because had massive audience.  Used this to create millions of email variations.  Created content repetition rules to surface categories that customers hadn’t engaged with before. 

– predictive AI enables deciding what to do next (best product, offer, channel) ; generative AI to match experiences to content or generate

– starting to see large language models and agentic AI in addition to personas and target audience segments in preferred channels

– challenge to centralize content, which is now often managed in last mile delivery system (email, CMS, website); large language models may be able to do this more centrally when connected to a good data set.

Personalization measurement

– measuring personalization requires bringing together all signals across digital, call center, and other channels.

– success metrics: one approach is personalization index being pushed Harvard Business Review and related to growth.  Want to get beyond opens and clicks to incremental impact on long-term measures such as customer engagement and protecting clients from competitive advances.  Measure how long customers stay active, how much they spend, attach items to shopping experience, visit extra channel, as well as over-all loyalty and lifetime value. 

– Incremental impact shows business value of personalization.  It requires marketing science using control groups. 

– need to look at long term, such as retention and lifetime value, which requires getting all data together in a reporting view, including offline data.

– can also look at engagement by channel and by segment across channels, which can be easier than lifetime value by individual and may be available in analytical tools provided by CDPs.

Role of CDP

– CDP lets act in real time, including data collection, recognition, sending personalized message immediately

– need a CDP to see data and activate across all channels. 

Data requirements

– first party data is critical to effective personalization. 

– data includes not just purchases, but also responses, media, clicks, and retail/offline experiences.

– a lot of first party data is in delivery systems such as email and website, but increasingly also in business systems, which can be offline.  Need both online and offline data to understand complete customer journey and do personalization.  A lot of needed data is already in a company data warehouse.

– need all company data such as product and store information that’s not in a classical CDP, for example store layout and inventory data so can personalize in-store messages to customers based on what they are predicted to want and what’s available.

– are applications such as inventory forecasting that can use company data and LLM which have nothing to do with customer data.

Data collection

– must have value exchange with client to collect personal data. 

– Need customers to trust they can provide email address without being bombarded with spam

– At Target, Brian used in-store locator to provide assistance while shopping, which is value to customer; data collected was useful to store to build deeper understanding of customer e.g. if spent time in pet food aisle then infer they are interested in pets and incorporate that within messaging

APAC Roundtable – October 24, 2024

Guest: Alex Burton, ANX Country Manager, Verticurl, part of OgilvyOne

Role of CDP

– lots of myths about what CDP does

  • most in the group felt that CDP is interest and intent machine to segment and bring unified data together; doesn’t do anything else
  • many vendors feel buyers want broader system including orchestration 

– common expectations:

  • single view of customer (look up all information about single individual).  This is largely delivered by CDPs.
  • journey orchestration across multiple channels.  CDP has data to enable that but it requires other technology as well. 
  • attribution and reporting:
    • expected since CDP does media and gets signals back, expect CDP to provide what they are now getting from web analytics and reporting platforms
    • CDP can make data available to reporting and analytics platforms but not replace them
    • distinguish basic reporting information on what CDP is executing, performance and customer behavior, vs. attribution, predictive modeling and optimization. 
    • Seeing more effort by vendors to integrate attribution, modeling and optimization into CDP but not yet wholly embedded. 

– unrecognized capabilities

  • many people don’t realize that CDP can also work with third party data and unknown as well as known customers
  • CDP is becoming more of a decision engine.  Well positioned because it has all the data.  Some execution engines trying to replicate CDP functions.

CDP Selection

– different CDPs do different things, have different strengths

– even if you buy a broad-function CDP, should still look for ability to integrate other systems as needed

– CDP marketing automation functions may meet needs of marketing team but not digital team.

– should be sure to look at multiple use cases when buy CDP, recognize that use cases will change over time

– right choice also depends on client maturity in terms of digital transformation internal skills and resources to support multiple systems

– a small organization with many different applications might be better off with replacing those applications with a single system such as Zoho One, which does most of what a CDP provides

Real Time

– real time can derail conversations if people don’t make distinctions and are impractical.  Need to frame the requirements more clearly.  Cost and nature are not well understood.

– types of real time capabilities:

  • real time access to individual customer profile, e.g. from call center agent or website personalization;
  • real time event triggers, responding to input data stream, e.g. scan for dropped shopping carts so can react with a popup message on the website; different from sending dropped shopping cart email within 24 hours
  • real time data ingestion: much harder than access or triggers.  Different types include:

– posting data into the CDP datastore as it’s received.  Already happens in many CDPs.

– processing the data such as calculating a predictive model score that’s immediately available

– updating the identity graph with new identifiers as they are received, and potentially rebuilding profiles if you connect two previously separate profiles or split one existing profile

– not everyone needs real time.  For many use cases, are other ways to do real time applications than CDP

– different activations require different speeds, could be subsecond or within one hour or four hours or 24 hours.

– real time processing is a very advanced use case, most people start with simpler use cases

– are often technical limitations in other systems that prevent real time applications. 

– source systems may only provide batch inputs

– delivery systems (“last mile real time”) may not be able to execute in real time

– can only go as fast as the slowest system in the process, e.g. may send list to Facebook but it just sits there for hours. 

– real time requires companies to think and act differently, so not just about using a new technology

Privacy

– big need for education around relation of CDP to consent management platform and other sorts of privacy systems

– CDP will not provide full consent management; those are separate systems. 

– CDP can work in areas such as consent compliance, based on data imported from consent management platform.

– CDP has great strengths outside of consent compliance, relating to privacy and respect of customer data.

– people don’t understand differences between privacy, governance and security. 

– much more interest in privacy in Australia today due to new law, but most Australian companies deferring dealing with the problem until the law comes into force in February 2025. 

– getting consent management in line with the law requires a lot of investment, but not much glory in running a personalization project.  Will have to wait until it becomes a board-level problem.

– companies should already be starting on privacy and consent practices such as data minimization and honoring preferences, even before regulation is ready.