Skip to content
Customer Data Strategy  •  Data

Ten building blocks for an actionable marketing data-asset – the Data Factory

 

Avaus has been active in sharing industry best practices and learnings in blog posts and webinars during the past years. To continue the tradition, I will guide you through the rest of the year, along with a Playbook of Avaus thoughts on how to build a future-proof marketing data-asset. Together with my colleagues, we will show how to utilise different data sources in ways that very few marketeers have yet applied in running their businesses.  

 

For sure you have heard the following phrase in many presentations and sales pitches: 

 

The right message, at the right time for the right person, perfectly personalised in the right channel”

 

Promises promises!

 

We bet that most companies don’t even get one of these things in place in a way that delights the customer, not to mention the cliché of the full marketing phrase…

 

Our recommendation is to select one item at a time and to start delivering with world-class standards. Then you will be ready to get the second item right following the same standards. In practice, getting two of the variables right at the same time means, in fact, an exponential increase in the level of complexity. We believe that building sustainable competitive advantage requires going back to start building the very basic right from the beginning. Artificial intelligence is not the first area in which to start investing!

 

Our Playbook for building an actionable marketing data-asset consists of ten steps or building blocks. Ideally, they should be taken in chronological order. We define an actionable marketing data-asset as the engine for timely relevant, personalised, multi-channel and cost-efficient sales, marketing and customer experience optimisation. 

When measuring relevance, being there at the right time with your communication can be more important than the actual content. That is the reason why being relevant is more difficult than tailoring content for a specific audience. To be timely relevant, marketeers need to be able to understand customer behaviour as it happens, not after two days of batch processing. It has to happen fast. 

 

Possibly everybody following industry trends has a large lake of data at least in the piloting phase, but is that data truly actionable? Do you have the needed data that reveals a customer’s intent? Is it in such good shape that it can be used in customer-facing communications? And, most importantly, do you have the needed consents to use all the data?

 

Building machinery to deliver timely relevant communication at scale is even harder. To avoid ending up with endless POCs and MVPs, right from the beginning you should always do the use cases and solutions at scale. Proven scalability is one key factor separating winners and losers in the present competitive environment. 

 

These are our suggested ten steps to build an actionable marketing data-asset:

 

 

1. The privacy-first approach

Beginning with privacy and regulation may sound boring. A privacy-first approach is crucial for any robust data-asset aiming to enable usage of new technology and use cases, especially since the aim is to combine different data sources for activation purposes. If the data is not intended for that purpose already at collections point, later usage will be limited. A privacy framework is unique for each company. 

The actual framework should follow privacy by design principles to enable, not disable. Privacy by design as terminology is already widely misunderstood. Based on our experience by using the methodology there isn’t anything  that couldn’t be done within the boundaries of not violating data subject rights or different legislations in disturbing ways. In the following post, we’ll guide through this process in more detail.

 

2. Start with simple use cases, but don’t forget the visionary ones

You start with an inventory that includes all your current activities and then you list your envisioned future ones. It is important to look far enough into the future to align the visionary activities being part of the privacy framework design not to slow down progress at later stages. Simple use cases usually deliver the most business value, so you should start with them until you run out of them. Only then you start to increase the complexity level. There will be upcoming blog posts & webinars focusing first on concrete tactics to gain efficiencies in digital media investments and followed by tactics to bring a marketing data-asset actionable in 1-to-1 channels besides digital media. 

 

3. Combining web behavioural data with customer data requires strict privacy compliance and technical enablers

To combine web behavioural data with other data sources you need to have the enablers in place, both in compliance and technical terms. Since being able to measure identified user behaviour across channels is one of the most crucial elements of an actionable marketing data-asset, you can already see why a privacy-first approach is required. In most cases, a valid legal ground to combine web behavioural data with other customer data sources is a consent defined in GDPR. Rarely can you rely on legitimate interest, at least from a data activation perspective. Therefore you need to have a consent mechanism in place, and this has both compliance and technical implications. 

Looking at previous first interpretations of the Data Protection Authorities across Europe, it’s inevitable that an informative cookie banner without an audit trail for a legally valid consent just won’t be enough.  Additionally, you need a common key to combine users online with known customers in an environment where typically cloud solutions are used instead of old premise data warehouses. Therefore PII-data elements or plain text customer identifiers need to be pseudonymised, and that needs to happen both real-time (request responding in 50ms)  and at scale. How to get all this in place? We are happy to share our learnings and mistakes too! 

 

4. (Web) Behavioural data usually has quality issues

There is a reason why the ability to collect, manipulate, understand and combine web behaviour data with other data sources has been raised to the top of the list. When compared, for example, to customer data, web behavioural data is usually not in good shape, as seen from a data quality perspective. Proper structure, naming conventions, continuity, coverage across digital assets and granularity are all missing. The collection of web behavioural data will be a rapidly evolving field in the coming years due to fundamental changes within cookies, browser privacy controls and general awareness of privacy concerns. For that reason, we see that it makes sense to build technology-agnostic solutions to enable web behavioural data being able to be collected whatever tools you are using. 

Practically, this means a fundamental shift in terms of organisational responsibilities, from web analyst to IT-development teams. In building actionable insights into a marketing data-asset, web behavioural data plays an important role by telling us what is happening right now, whereas customer data is more static. Thus we see that putting web behavioural data in good shape is an area of investment for the majority of companies, in order to secure future competitiveness. This topic, like all others, will undergo its own deep dive in future posts!

 

5. Segmentation is all about creating customer (prospect) profiles

Whilst at this stage web behavioural data for identified users is combined with relevant customer data for sales, marketing and customer experience purposes, the first not-so-boring step, at least from a business perspective, is making sense of all the new insights that can be created from this joint view. To be precise, two separate activities are  required for smart segmentation:

  1. Analysing behaviour on digital services at an identified user level
  2. Manually creating new target groups or audiences using combined behavioural and customer data

Using identified customer data to analyse online service behaviour opens doors to an infinite amount of possibilities to understand user behaviour significantly better, especially building hypotheses around the differences of behaviour between various types of users. For instance, how do high value vs. low-value customer behaviour differ, or which behavioural patterns can be detected before a customer churns? 

Unfortunately, new insights and analysis have little value besides pleasing the ever-growing appetite of management for reports and nice graphs on PowerPoint. Value capture begins when something is done with the newly acquired insight, namely in the context of sales, marketing and customer experience to build communication activities with clear targets. To follow the examples given, to build activities around low-value customers guiding them to behave like high-value customers, or clear churn prevention activities for those customers that show signs of potential churn. 

Since capturing value is mostly about execution, these activities can rely on manual processes at this stage to prove that earlier hypotheses are on the right track. Evidence should ideally be captured from a channel that is easiest and fastest to reach customers. One channel to consider at the outset is Search since budgets are usually at a level where a small impact already makes a big difference. No heavy integrations are required either. We’ll deep dive into this area also in future posts together with Oscar! 

 

6. Activate new target groups across all possible channels

After gaining your first new insights and building first target groups or audiences, it’s time to start looking into creating some new activities around the marketing data-asset  to prove its actionability. From a learning perspective, it’s wise to start activating new target groups or audiences based on customer (prospect) profiles consisting of multiple data sources, one channel at a time. Over time it’s essential to build capabilities for activation for all channels in use within a company ranging from digital advertising, website/login service personalisation to all 1-to-1 channels like e-mail and telemarketing. 

Depending on your overall architecture for marketing technology and other IT-systems, there are multiple ways to do this. Unfortunately, there is no one-size-fits-all recipe. However, some technology stacks have certain advantages and benefits, whereas being dependent on old on-premise architectures decreases time to market significantly. From a customer experience perspective, customers should be addressed in the channels that they choose, not the channels in which a marketeer hopes to reach them. The activation topic can be approached by some example scenarios where the ease and speed of data usability plays a key role. 

 

7. Bringing it all together into a real data-asset

It took us quite some steps to get to a point where we can really start talking about a concrete actionable marketing data-asset. In our definition, there are a few characteristics that uplift a collection of data points to an actionable marketing data-asset. Typically the following should be in place:

  • Relevant data points for sales, marketing and customer experience optimisation use cases organised around one customer (prospect) profile 
  • Global unique user identifier to combine multiple data sources
  • Data in format to be used easily for creating insights and target groups or audiences for activation
  • Legal basis definition and consent values as a native part of the customer (prospect) profile. Additionally an audit trail of consents or opt-outs in a centralised repository
  • Integrations to all channels used available even on a manual basis at the early stages of development
  • Governance model in place including aspects like privacy framework, access rights, data ownership, GDPR related activities like DPIAs etc.
  • Performance data available as a feedback loop on business value delivered for activities
  • Data retention rules for different data points and PII-data redaction mechanisms in place
  • Proper documentation that is continuously updated as changes are made to fulfil compliance requirements and ensure smooth collaboration between teams. This part of building any data products is almost always neglected and decreases productivity over time.

As noted earlier, marketing data is rarely treated as a data-asset and when looking at the list above, it’s clear that it takes some effort and investment too. Luckily there are fast ways to get started to fulfil easy-to-implement and high business impact use cases with low upfront costs. We hope that you have got a basic understanding of elements to get started and we are happy to elaborate on the topic further!

 

8. An algorithmic approach to activation

As previously mentioned, significant benefits can be achieved with simple use cases (from an analytical point of view) which are fast to implement and deliver measurable results. Having gained the easiest wins and paying the investments so far, it’s time to start looking at more complex cases. Within the past years, usage and out-of-the-box availability of algorithms have changed fundamentally, where the algorithmic approach is no longer a data science challenge of building models, it is rather a data and engineering challenge of cleaning, formatting and getting the data in a shape thatz can be used for modelling purposes.

The utilisation of algorithms is usually more of a data engineering and customer-facing process implementation challenge, than actually building models providing high predictability. Whilst building an actionable marketing data-asset, the data engineering challenge has already mostly been solved due to having data points for sales, marketing and customer experience optimisation purposes in a nice clean format organised around the customer (prospect) profile. As a consequence, the utilisation of algorithms will be easier and faster from a technical perspective by getting the data-asset in place first. 

This doesn’t mean that using algorithms would be easy at all, you need to have matured enough on customer-facing and business processes to put them into production use at scale. Finding enough support on the management buy-in level is also key to achieving tangible results in a company algo leap. Avaus has lately been focusing heavily on building a specific data x algo x action approach which we’ll share more insight on to get your company truly algorithmic driven.

 

9. Automate the activation and execution 

Especially when entering an algorithmic method of creating activities, there is a growing need for automation within triggering activities, deciding upon prioritisation, getting the right content in place and optimal timing of the activities. Personalisation can only be accomplished using at least some automation since it’s impossible to handle it manually due to the high level of variety in the content used for individual profiles. Once you have enabled the marketeers to work with new types of insights, target groups or audiences, activating them across all your channels in cost-efficient ways needs smart automation to connect to various APIs. 

When building this capability, it makes sense to plan it right from the beginning in a multi-channel context. As the approach towards target groups and audiences starts to get sliced over time into smaller but contextually more relevant slices, proper automation is already crucial to the execution. To set the vision to guide you to the future, there should never be communications without follow-up actions, whether the user has reacted to them or not. 

Moving from data engineering-focused tasks towards activation at scale will require different skill sets from the team. Development skills to execute and creative thinking are essential when increasing the amount of new and parallel activities. In these times, when we are living with pressure to keep businesses running, humans shouldn’t do anything machines can, both from an efficiency perspective and to keep humans happy occupied with interesting non-routine tasks. 

 

10. Orchestration of activities

The most challenging part of utilising an actionable marketing data-asset is the orchestration of different activities. Orchestration acts as a decision engine to put the customer on the right journey, to decide on a sequence of activities or channel preferences. It also enables you to act with various activities across all of your channels in a consistent and continuous way, creating delightful or even surprising customer experiences. It’s certainly an area where there is still more marketing hype than practical experience across marketeers. Architectural design for marketing technology components and operational data accessibility will, in the end, define the boundaries for orchestration. 

 

Putting myself in the architect position to design capabilities for communication orchestration, I would thoroughly consider which are the key capabilities that can be provided by commercial software and which enabler would be crucial to even develop in-house to avoid vendor lock-in and to provide flexibility in an area where technological development is faster than most of us can keep an eye on. Despite being in the very early stages of customer communication orchestration, I’m sure there are plenty of best practices to be shared and further discussed with you.

I hope you have managed to read until this point and gained some knowledge to plan your organisation development for better utilisation of data for sales, marketing and customer experience optimisation. At the same time, we are inviting you to join the journey together with us to take marketing data-assets to the next level.

 

AUTHOR

Teemu Relander

Head of Data & Analytics
teemu.relander@avaus.com