Eliot Harper defines what a CDP is (and is not) and shares five traits marketers should look for in a CDP.

CDP (customer data platform) has emerged, almost overnight, as yet another must-have TLA (or ‘three letter acronym’) that every marketing team simply cannot live without. CDP industry revenue hit $740 million in 2018, up 50% from the previous year, and is estimated to exceed $1 billion in 2019, according to the CDP Institute. And the number of CDPs is picking up pace. At the recent B2C Salesforce Connections event in Chicago, Salesforce announced it is launching a CDP, then on the same day, Adobe announced its own one too. Not to be outdone, Oracle also chimed in and vaguely explained that it has enlisted the help of Accenture and Capgemini to help them figure out CDP.

But while marketers are busy talking about CDPs and vendors are scrambling to build them, there’s still much ambiguity concerning what a CDP actually is. That’s largely due to the fact that there are conflicting definitions for the acronym, several of which have been authored by software vendors. Just like the Cinderella fairytale, several vendors have made a vain attempt to shoehorn their data platforms into the ‘CDP glass slipper’ in the hope of a triumphant reward.

But without an unambiguous explanation, it’s hard to determine what benefit a CDP provides, or why you would actually need one. So in this article, I’ve analysed definitions, deciphered vendor claims and assessed current platforms, to arrive at a definitive interpretation of what a CDP really is and what you should expect from one.

 

CDP versus CRM

Before defining what a CDP is, it’s important to understand what it is not. For starters, a CDP is not a customer relationship management (CRM) platform. You would be forgiven for assuming that it is, as they both store customer data, but that’s where the similarities end.

By design, a CRM is built for storing known customer data, which must conform to a rigid schema. A CDP, however, derives customer data from disparate sources, including first and third-party data, structured data, semi-structured data, unstructured data, website and mobile app session data, and more. And this data needs to be updated and accessible in real time. Sounds reasonable? Just ask your IT team for all that and let me know how you get on.

 

CDP versus DMP

A data management platform, or DMP, also bears the hallmarks of a CDP, as both platforms use unknown data. However, a CDP is designed to work with both known data for an individual (for example, contact and transactional data) and unknown data (for example, website cookies, IP addresses and mobile device IDs). And a CDP consumes this data from many different sources, including data warehouses, data lakes, operational databases and other disparate sources. A DMP, however, is largely designed to work exclusively with unknown data from digital channels and advertising networks.

 

CDP versus ESP

Enterprise-grade email service providers (ESPs) have evolved from bulk email platforms into inclusive digital marketing platforms that nurture customer lifecycles and automate repetitive manual processes, across multiple channels. But most vendors distance themselves from the ESP label, for fear that it positions them as a glorified email cannon.

New martech acronyms have also popped up along the way, including marketing automation platform (MAP), integrated marketing hub (IMH) and multi-channel marketing hub (MMH), but none have stuck. At their core, these platforms are still a database marketing platform, optimised for delivery of relevant marketing messages, primarily by email. Yes, they can do more than send emails, but you’d never buy one of these platforms if you weren’t going to send emails. So we’ll continue to borrow the acronym ESP for a little longer, until a better one comes along.

If an ESP uses marketing data for, well, marketing, isn’t a CDP actually an evolution of an ESP? Not exactly.

While most ESP platforms can perform extract, transform, load (ETL) processes on data from external sources, they can only run as scheduled batch processes. And by design, they heavily rely on a notion of a ‘Subscriber’ (or a similar noun) to represent a known individual (who has a name and an email address), not an unknown one. And if an individual is unknown, then they simply don’t exist.

 

Big five personality traits

In psychology, personality is typically modelled on five broad traits: extraversion, agreeableness, openness, conscientiousness and neuroticism. In a similar way, CDPs should also exhibit five common personality traits: data integration, segmentation, engagement, data orchestration and actionable reporting.

1. Data integration

Data integration is where a CDP makes the impossible possible by enabling end users (marketers) to actually be end users, allowing them to configure data integration, quickly, without an over-reliance on IT. The key is to integrate every data artefact related to an individual, and this is where you need to pay close attention to what it says on the box. A CDP should offer out-of-the-box connectors for different data sources – the ones you know you need today and others that you don’t yet know you need. Without this, you’ll be stuck with an incomplete view of an individual, at best.

Data integration types can be divided into three categories: structured, semi-structured and unstructured data.

• Structured data includes SQL databases that neatly conform to a relational schema. Database platforms include Amazon Redshift, Snowflake, PostgreSQL, MySQL, Google BigQuery and Microsoft SQL Server.
• Semi-structured data is typically formatted as a delimited text file that doesn’t conform to a defined schema, but which has properties that make it easy to analyse (for example, column headings).
• Unstructured data is other data that doesn’t really conform to a schema. This includes streaming data from website engagement implemented by tag-management systems (like Google Tag Manager), and siloed data from streaming data platforms  such as mParticle and Segment.

And not only does a CDP integrate disparate data sources, it also uses complex matching algorithms to unify identifiers from the different datasets to create a single customer view that truly understands every aspect of your customer – something that has been the Holy Grail of marketing for over three decades and has been shunted into the too-hard basket, until now.

2. Segmentation

Simply put, segmentation allows you to identify a group of people, or ‘audience,’ to target in your marketing programs. The term can be broadly distributed across four categories:

1. Demographic, which includes age, gender, education, occupation, family size and more.
2. Psychographic, for lifestyle, interests, preferences, social status and personality traits.
3. Behavioural, including engagement (for example, website and email), purchasing habits, loyalty, website usage and purchasing frequency.
4. Geographic, based on continent, country, city or local region. 

Most ESPs already offer segmentation, which generally falls into two groups: filter-based and query-based.

With filter-based segmentation, platforms typically offer a logical expression editor based on ‘AND OR’ statements. For example, you might choose to segment Contacts that are based in Europe AND have an email address OR a mobile number. While such a segmentation approach is easy to use and effective, it’s also very basic and very restrictive.

Query-based segmentation provides the ability to use database query languages like SQL, where a database administrator can literally knock themselves out and build complex segmentation criteria. The problem here is that while SQL is an extensive declarative language, it’s not the lingua franca of marketers. It’s a specialist discipline that not only requires familiarity with the language’s vocabulary but also knowledge of optimised database design and query design patterns. Oh, and you need to wait for the data to finish processing before you can use it – forget about getting this data in real time, or even near time. Typically, complex queries run sequentially as a long linear chain. You need to wait for the process to start and then end, which can take several minutes, sometimes hours.

A CDP, however, provides the ability to build complex segmentation using a ‘clicks not code’ approach. When it comes to segmentation, a CDP should, at a minimum, tick the following six boxes:

1. Easily define your audience, without the constraints of classic filter-based user interfaces.
2. Immediately see insights (like audience size) after segmentation.
3. See how audiences overlap.
4. Create waterfall audiences (audiences with two or more mutually exclusive segments, where a customer is only in one segment at a moment in time).
5. Allow dynamic segmentation, where an audience is continuously updated whenever data changes.
6. Create test groups, based on a randomly selected percentage of your audience.

These six points are critical in segmentation today, as there has been a shift from traditional funnel-based marketing. Individuals no longer qualify to join an audience based on a logical test. Instead, in the new marketing funnel, leads can enter at any stage and segments need to be highly pliable.

3. Engagement

Engagement refers to the ability to trigger an action based on a real-time event; for example, a customer installs an app on their mobile device, or a prospect adds a product to their ecommerce cart and then leaves the website. When these types of events occur, it’s often an optimal time to engage the individuals involved and direct them towards an outcome; for example, send them an email or SMS message.

And engagement is the point of demarcation between a CDP and an ESP. While the unique needs and desires of prospects and customers are determined through a CDP and the single customer view that it provides, audiences then need to be engaged through marketing communication. Engagement is the baton in a relay race, and also the point at which many CDPs lose their grip.

The first consideration is that you need to trigger the action immediately, not after a pregnant pause. To achieve this, a CDP should provide application programming interface (API)–based integration with your ESP, in order to inject individuals into journeys or trigger messages in real time. Don’t compromise for anything less.

Another consideration that many CDP vendors fail to comprehend is that an ESP requires more than an audience dataset to send a message. You not only need to know who to communicate with, but also what content to communicate to them. For example, if a customer has recently made a purchase, then you might want to engage them by sending an email that lists the items they ordered and solicit feedback. The problem is, content personalisation is a tricky business. ESPs typically rely on AMPscript, Emarsys Scripting Language (ESL), Responsys Personalisation Language (RPL) and other proprietary scripting languages to personalise content. You can’t personalise content just with data attributes alone – well, at least not to any sophisticated level. This is where many CDPs seemingly have a ‘It’s not my problem’ mindset. They provide the data, then leave it to the marketer to figure it out.

In an ideal world, a CDP empowers marketers with a content payload of programmable content, designed for the target platform, that marketers can insert into messages, without needing developers to get in the way.

4. Data orchestration

Data orchestration refers to the notion of a harmonious coordination of all data across multiple platforms, to make your data sing. Remember that a CDP is just one component within a broader technology stack. It needs to tightly integrate your CRM, ESP and data platforms. Only then will you be able to create a real-time single-customer view.

There are two key principles in data orchestration. The first is to maintain a CRM System of Record. A CDP should not only import lead or contact records and their associated attributes from a CRM, but also maintain the System of Record. If a lead record converts to a contact in a CRM, then the CDP should be aware that the conversion has occurred. And if multiple contact records are merged into one, then the CDP also needs to be ‘merge-aware’, to ensure record integrity.

Secondly, a CDP should round-trip data; in addition to triggering journeys or messages, it also should ingest engagement data back from the ESP. Email engagement data like opens, bounces and click-throughs is an important slice of a single-customer view. And a CDP should unify this engagement data with other event data; for example, from a tag management system, to identify website interactions from the click-through breadcrumb trail and whether the individual ultimately converted.

5. Actionable reporting

With all this data flowing in from multiple sources, you need to be able to use it, which is where actionable reporting comes into play. Data needs to be not only reportable but also actionable, and you won’t find both of these words in every CDP vocabulary. Look at Datorama as an example. While this platform is not a CDP, and does not pretend to be one, it does set the benchmark for marketing data reporting. This platform provides market intelligence reporting by centralising data metrics from disparate sources into one consolidated interface, making it really easy for marketers to optimise their marketing efforts. This is all very nice, but it’s not actionable. While you can gain powerful insights regarding the effectiveness of your marketing spend, you can’t act on that data, triggering communication to specific audiences like unconverted leads or unengaged contacts. At least, not yet.

 

Separating the pretenders from the contenders

CDPs are continually evolving, and while not all platforms exhibit these five traits today, they may do so in the future. (One contender that does exhibit them right now is Stride Software. If you’re not already aware of Stride, then you probably should be.)

Without these five traits, a CDP will only compromise the value that it sets out to provide. But with them, a CDP is a compelling tool for marketers that promises significant returns.

Several CDPs have exactly this paradigm: you can see the insights but you can’t action them. The intention of this article isn’t to name and shame CDP vendors, though – they know who they are, so I’ll leave it right there.

 

Eliot Harper is a freelance marketing cloud expert.

 

 

Image credit margouillat © 123RF