Fireside Chat #2 – Customer Data Platforms, Worth It?

Topics Covered

1. What exactly is a Customer Data Platform and what is it supposed to do?
2. Why is it that there has been an explosion of ‘CDPs’ on the market over the last few years?
3. Are CDPs all the same under the interface?
4. Under what conditions would a CDP make sense to a business?
5. Would we recommend different routes depending on various internal conditions and requirements?
6. What is your experience of how much value CDPs actually deliver after implementation?
7. Why is it that CDPs seem to stall in terms of delivering value after an implementation?
8. What would a good delivery framework look like to actually achieve the value return it should do from a CDP?
9. Do we have any resources that can help businesses learn from our experiences?

Answers

1. The goal of a Customer Data Platform is to create rich unified profiles of customers and make that available for insights and activation purposes primarily to support marketing and sales activity in the context of better and more relevant customer experiences.

From a business perspective it should provide a much richer dataset so that they can have a deeper understanding of the customer, more accurate measurement of engagement at a customer level and provide a tool to more easily attempt cross channel marketing and personalisation activity.

From a customer perspective it should result in them having a much more relevant and consistent experience with a brand’s touchpoints which is going to increase a brand’s customer retention and lifetime value.

2. I think there are many reasons why there has been an explosion of CDPs but I would say that some of the key ones are:

• • Legal and regulatory changes aimed at providing better controls of privacy for consumers have pushed brands to consider what a first party data strategy should look like and how they can achieve it with existing technology. It’s been more of an issue in Europe where stricter laws and regulations have been introduced but I believe that brands globally should be considering this too
  • A huge increase in the different data and marketing platforms available creating more siloed data sets and more touchpoints for marketers to consider. I think even the latest number of CDP vendors identified by the CDP Institute was over 160, the last time I checked and I can believe there are hundreds of tools available in every single data and marketing category. This obviously creates the need to find tools that can easily manage to pull these data sets together or orchestrate experiences from a central system.
  • Technology advancements have fueled an increase in consumer expectations for personalised experiences. Personalisation technology is becoming cheaper and more accessible for brands to have advanced capabilities that 10 years ago were not widely used. Now relatively cheap tech is available to everyone so they can achieve highly personalised experiences which in turn has increased the customers’ expectation for these experiences. Therefore brands are looking to CDPs to help support personalisation to be able to compete.

3. If you listen to any sales pitch or read marketing documentation they all position themselves as being capable of doing the same thing but in reality there are many differences between vendors, even relating to core functionality you’d expect from a CDP.

The 3 core features of a CDP are that they can ingest data from multiple data sources, they can stitch this data based on customer identifiers to create unified profiles and they can make this data available to external systems for analysis and activation.

The more established vendors seen as leaders in the industry can do all of the above. They’re able to ingest large volumes of data via streaming or batch methods, unifying profiles based on various customer identifiers and have large libraries of out of the box integrations with downstream systems or other export methods where custom integrations can be built. The slight differences that we do see across some vendors relating to these core features are:

• • Timing related to the ingestion and processing of data where some have significant delays from where data is initially made available to the system to it actually being accessible for insights or activation. These delays can then be further compounded when the external activations have additional delays which can then start to restrict the scope of activities you may be able to design with the system
  • Stitching process which includes a few factors:
    • Whether the system can carry out a matching process deterministically, where you are only using a direct match if identifiers exist in both data sets, and probabilistically (fuzzy matches), where the system is joining records based on multiple data points to try to stitch records where exact unique matches do not exist. For vendors that can carry out probabilistic matching there is then a difference on the flexibility and control you may have on this matching process. Many vendors we speak with only match deterministically but then offer integrations to external systems that provide data sets to allow a client to carry out probabilistic matching.
    • How many identities can you stitch with? Many vendors can carry out the stitching process on multiple identities but we see a lot of email service providers, experimentation and analytics tools that also market themselves as CDPs but actually only unify profiles based on a single identity. This is definitely a useful feature but again would significantly limit use cases that you would want to attempt to drive from a CDP.
    • Handling of grouped or hierarchical profiles so that it is possible to report and activate at both a group profile level of which there could be many levels within a hierarchical structure and an individual profile level
  • The out of the box integrations available for external systems for analysis or activation purposes. Some vendors have large libraries of integrations available that allow for easy configuration of starting to push out data to these systems whereas many require custom integration work to be carried out which is technically possible but adds to the investment in time and resource to get use cases delivered across platforms where this is needed. It also means that these integrations will need to be maintained by the CDP customer rather than relying on the external system vendors to carry out updates and testing, all adding more resourcing requirements

Outside of these core features I would say other features that differentiate vendors relate to data governance, journey orchestration, ETL functions, ML/AI predictive capabilities, reporting and insights and experience delivery. It really does vary across the CDP landscape and this is because the industry is full of platforms that have origins in other areas and have added to their features.

4. I think the conditions that make sense for a business to consider investing in a CDP are where they have:

• Multiple separate customer datasets but want to report on data at a customer level
• They are trying to engage with customers across multiple touchpoints
• They have teams with a desire to optimise their own channels activities using more data; and
• They have internal resources that can lead a cross team initiative

Most of the issues we see CDPs resolve are issues caused by silos of customer data and the inability to easily create cross channel journeys and report on these. Some examples of the issues we see are:

• Customers having suboptimal experiences by being targeted with content that is irrelevant or of low relevance due to personalisation attempts based on data with gaps in it. For example, dynamic retargeting of products in paid media might be based on products previously purchased but refund data hasn’t been considered and therefore ads may be trying to upsell or cross sell other products based on products that a customer doesn’t actually have an interest in.
  For B2B clients, we’ve seen scenarios where sales and accounts teams are having initial conversations with customers based on no prior knowledge of the customers interests or limited information included in a form submission. These conversations could be greatly improved if, for example, they had access to summarised information about a customers affinity for particular products in a category based on onsite and offsite behaviour
• Inconsistent messaging across channels where, for example, you’ve booked a holiday or bought something and then while receiving emails to confirm this you are then still constantly targeted by advertising for the same thing I’ve just bought and potentially emailed also about that too!
• Confusion and mistrust internally on performance data where reporting is inconsistent. Again, one thing we commonly see is channel performance reporting differ from media platforms vs reporting from digital analytics tools vs BI systems particularly when trying to understand this at a customer level causing internal friction.
• Large investments in time and effort to coordinate cross channel campaigns across completely separate systems in each team which ends up in teams sticking to BAU activity with few cross channel considerations and experimenting or improving their engagement with customers.

5. I think some of the key questions that we consider when determining what route to take with a customer for investing in a CDP or if at all are:

• What is the state of existing data sets that we want to join and use for insights and activation. For example, we recommend onsite data collection uses a data layer for robustness and accuracy and then, if it is implemented, does it contain a rich set of data points to provide the context you need to understand customer engagement with products, services and content. Also whether its products, services or content personalisation that a customer wants to implement, these all require assets and data points to feed dynamic experiences, which can only be as good as the quality of this data powering those experiences. Finally what identifiers are available to identify and join customer data across systems.
• Are clients maximising the existing technology they already have? This is particularly important where we see clients that have invested in large technology eco systems such as with Adobe or Salesforce but are potentially not yet maximising the features and integrations that already exist that allows them to attempt some cross channel activities.
• What is the complexity of targeting and personalisation activity currently being carried out? Buying a CDP will not automatically unlock the internal process, expertise, assets and data required to achieve these things. Where businesses are carrying out some of this activity, there is a good foundation to build up on with the CDP
• What are internal resources to drive CDP activity? Having internal teams that are dedicated or able to build out the time, processes and skills from a business and technical perspective to focus on CDP driven activity is essential. As with all new investments, it is possible to own all of the tech in the world but if there aren’t the right teams to deliver then nothing of value will be achieved.

6. This ultimately depends on the customer’s use of data that is feeding their activation platforms and insights activities. But potentially uncapped value could be seen if the right team is maximising CDP data and features. Identifying what unique opportunities are available with access to fully enriched profiles and easier orchestration of cross channel journeys will help to focus on what unique value the CDP provides.

To initially see some value from the CDP, vendors will often suggest paid media suppression and retargeting use cases as these are extremely easy to implement and should immediately improve ROAS and ROI metrics for paid media channels. For clients with large media budgets these efficiencies and improved retargeting capabilities can quickly partially cover the initial investment of a CDP.

Focusing on high and medium impact use cases with low to medium effort and most importantly not trying to attempt overly complicated activity. A simple example would be where a customer may gain access to offline refund and cancellation data which in the CDP is merged with tracked online behaviour, the email and onsite teams can provide product recommendations that reflect more accurately the customer’s engagement considering the offline data and this should improve performance of those recommendations. This can easily be rolled out before trying to incorporate more complex predictive scoring data to feed recommendations. Another example is looking at how paid ads can support owned channels such as email where a team can look to re-engage an abandoned customer through “free” methods before potentially paying to bring them back again.

7. There are several factors that we see stall value being delivered from a CDP:

• Lack of clarity and long term use case planning which is due to the complexity of CDP projects which require so many teams, multiple data sources, masses of activation channels and a broad spectrum of skill sets. Requirements tend to quickly snowball and we’ve seen customers frequently attempt to plan impressive cross channel activity and advanced dynamic personalisation but the requirements become so complex, that the mammoth efforts are required for activities to go live and eventually nothing goes live.
• Data sources aren’t actually ready to be used due to inconsistency, errors and gaps in data sets. Where this occurs, significant delays are added due to effort required for internal teams to rectify these data issues. This can be really significant where we’ve seen customers who have prioritised onsite activation experiences, such as implementing product recommendations modules driven by algorithms using onsite and offline customer data, but onsite they do not have a comprehensive or robustly developed data layer. This then requires technical design and development tasks to be carried out prior to enabling more robust onsite data collection.
• Internally alignment relating to business KPIs and expected timelines for ROI. A misalignment in goals and timelines set by boards and C-suite who have approved investment but expect specific performance levels in unrealistic time frames can also result in teams not investing in a CDP as much as required. Internal teams are forced to continue focusing on BAU activity or rely on discounting and promotions instead of building out more complex activities using new personalisation capabilities based on CDP data because these old methods have delivered expected performance (at least superficially and unfortunately are not sustainable in the long term).
• A lack of expertise. There is no point in having the technology if there isn’t the internal resource to maximise the functionality of the tool. You can have all of the data and tools in the world but if teams aren’t using them or know what’s possible then no value will be achieved. It requires full engagement from resources that are designing, implementing and maintaining requirements of the CDP built from a long term strategic roadmap. This all needs to then be supported by a skilled team for the experience delivery and insights to continue identifying opportunities and measuring success. It isn’t an uncommon scenario we’ve seen where CDP customers have rolled out initial use cases but then no individual or team has taken ownership of the CDP. Therefore no long term strategic roadmap of activity is developed to maximise ROI and becomes another piece of unused platform in an ever growing stack of redundant tech.

8. The 4 key pillars that we consider for developing a good framework when working with a CDP (and any technology solution really) are:

• Data primarily considerations for data points that will feed customer segmentation and personalisation capabilities
• Technology for features that determine the scope of capabilities plus the technology that will actually deliver experience
• People (possibly actually the most important) is the team with the capabilities to build an audience and journey orchestration strategy, technically implement requirements in the CDP, resources to build out the experiences in the downstream systems and analytical resources to identify opportunities in the CDP data
• Processes in order to develop a testing culture where requirements and experiences are delivered in an iterative way learning from the data collected from past delivery. Also as importantly are the processes to enable cross team planning, communications and development of experiences

9. I’ve spent a lot of time working on CDP projects, supporting RFI/RFP processes and doing research on CDP platforms so that we can understand the full feature set, strengths and limits of these vendors. Off the back of this we have scoring and summaries across 8 or 9 vendors that we more regularly engage with which we’re happy to discuss with anyone.

We’re planning to release a whitepaper over the next couple of months with this information in it to help people start to differentiate vendors more

I have also written a series of blog posts that look at the importance of data validation, use case planning, id unification and how to assess vendors which you can find on dmpgteam.com.

The last thing we have is a full set of questions to be used in an RFI process covering all key areas of CDPs from data ingestion, processing, storage and governance to profile, audience, insights and segmentation capabilities. If this is of interest feel free to contact us directly and we’re more than happy to share this and discuss anything in the CDP world.