Approach to data science transformation

This video shows you how to transform an organization that has implemented or is implementing Pega Customer Decision Hub™ (CDH) in terms of Data Science.

 

Data Science Transformation is a shift to a broad adoption of Machine Learning and Data Science practices, to enhance and optimize the customer experience across various touchpoints and interactions with an organization.

 

In the context of Pega Customer Decision Hub, Data Science Transformation refers to:

• Adoption of the CDH Machine Learning capabilities (Adaptive and Predictive)
• Continuous optimization of predictions and models
• Enablement of Data Science resources
• Establishing governance for MLOPS

So why should we focus on this topic?

 

By harnessing the power of Machine Learning, organizations can unlock the full potential of Pega Customer Decision Hub and deliver, at scale, exceptional personalized customer experiences that meet and exceed customer expectations.

 

Data is a central asset in Customer Decision Hub. Data Science Transformation enables organizations to extract valuable insights from the data and convert them into actionable strategies in real-time by using Adaptive and Predictive Models.

Pega Adaptive Models predict a customer's propensity to interact with messages, so you can select the right one at each touchpoint to make every customer interaction feel relevant and personalized. Adoption of Adaptive Models is key to a successful Customer Decision Hub implementation.

An essential foundation to transformation is identifying key stakeholders from the organization to take on the journey.

 

Key stakeholders in this area are those responsible for the data science function in the organization, in other words, data scientists and model governance teams. The approach to transformation will vary based on the presence of a data science function and its maturity.

 

A potential challenge is that there may be a data science function, but one that has no current, direct involvement with marketing or customer communications. In this case, strategic direction must be discussed at a senior level to make a decision about whether to get such a team involved with marketing going forward, form a new team, or work as if there was no data science function.

Another challenge can be the definition of data science. Some organizations might consider themselves to be working in this area, but in fact are working on more traditional data analytics. It is important to understand the day-to-day workings of the identified data scientists or data science teams, to verify that the nature of their work is aligned.

Whether they have a data science capability or not, key stakeholders in this area will always include reporting and business intelligence teams, data and business intelligence operations, and MarTech (the name for all the different types of software that marketers use to optimize their work). Indirectly involved, but very important to this area are marketing ops, marketing strategy, and business sponsors.

As well as identifying the roles that fulfill the specific functions in which we are interested, it is worth reviewing and reflecting on information about the organization as a whole to ensure that all potential stakeholders are considered in context when assessing the current situation, and then applying typical stakeholder analysis methods.

Now we have discussed who in the organization needs to be involved, it is important to consider the medium of interactions with the client.

 

For major workshops to establish the current business situation and collaborative workshops to design a future state, working in person with the client can be great, especially to establish the relationships early on and make detailed observations and assessments of stakeholders and the business situation. However, there are plenty of tools out there to help you conduct these types of workshops remotely and still keep them lively and interactive, for example, Mural.

 

If you need or want to conduct such workshops remotely, it is worth having a brief in-person meeting with key stakeholders beforehand, to introduce yourself, set expectations, and understand their point of view. This will make the workshop much easier to facilitate.

When considering interactions, always bear in mind your foundational stakeholder analysis. For example, some very senior stakeholders may appreciate a one-to-one briefing before and after a workshop, rather than directly taking part.

When would we tackle these topics during a typical MLP delivery?

 

This timeline gives us a typical timeframe for an MLP, but of course, depending on the scenario, this can be done faster. We recommend tackling the organizational transformation topics in a specific order, based on best practice and field experience.

 

As you can see, this topic is addressed following the foundational topics of Mindset Transformation, Governance, Organization and Agility, and KPI, Reporting and Value Measurement. At this point, you should have a really detailed view of how the organization works and a good relationship with key stakeholders, making it easier to jump into the detail of this topic.

During dedicated meetings in this period, consultants should review the current data science capability within the organization to establish the best way to enable them on machine learning in Pega Customer Decision Hub. This ensures optimal integration and adoption.

So what are the key tasks that need to be undertaken in these areas?

Pega has a standard MLP backlog when delivering CDH, and under the adoption category, we have recorded all of the tasks required for organizational transformation.

 

For each of these areas, we have a common approach to achieving transformation:

Step 1: We come to a mutual understanding of the current business situation as it stands right now, regarding the current topic. How does the organization currently use any machine learning, how mature is the data science function, what are the governance processes that are used around models, what is their current view and understanding of machine learning in Pega, and how do they understand machine learning as it fits into their organization?

Step 2: We present best practices on CDH machine learning and model governance derived from years of global implementation experience. We will cover these best practices in the next topic and how it should be adapted based on the current situation.

Step 3: We define client-specific data science processes for CDH, including integrating external models, and maintenance, optimization, and governance of adaptive models. We keep as close to best practice as we can.

Step 4: We then define a plan that includes steps to get from the current state to the mutually agreed upon to-be state, to really bring it to life and make it happen.

These four key steps to success are what we feature in the MLP backlog.

[How]
Now, what are the recommended methods to use to undertake these tasks?

 

As explained earlier, Pega takes a common approach to tackling each of these topics, so let's start with our common step 1, which is to review the current state. Pega recommends that you apply as many of the following techniques as you can, to establish a clear picture of the current business situation:

 

• The most useful technique will be a workshop, or workshops, where key stakeholders present their current ways of working to you and discussions can be conducted. For this, you must set expectations in advance so that participants come prepared. A workshop is invaluable, first of all, to establish your own relationship with the client, but also to observe and understand the dynamics between the stakeholders and to refresh your original stakeholder analysis, if required.
• The workshop approach will help you gain a common understanding of the situation as the client understands it, with any inconsistencies or disagreements likely to be aired and resolved as the client works together to explain it to you. This is the primary advantage of workshopping rather than just talking to individual stakeholders. As a facilitator, constantly play back your understanding by either using diagrams or visual notes. Also, be careful to create space for all stakeholders to contribute. If there are some quieter stakeholders or some dominant characters, you can do simple things such as asking people to note their thoughts on a matter individually, and then you as a facilitator take those notes, read them aloud, and create a diagram to illustrate a unified picture. It can also be useful to make note of quieter stakeholders and find another way to engage with them after the workshop.
• A workshop will normally be the first step in reviewing the broader picture. During the workshop, note documents mentioned in the process so you can perform document analysis afterward and add to your understanding.
• Also during the workshop, make note of individual stakeholders with which it would be useful to perform protocol analysis, shadowing, or conduct interviews with after the workshop to really build that complete picture.
• Once all investigative techniques have been exhausted, be sure to clearly document your understanding and play it back to key stakeholders for some sort of sign-off or agreement. This is an essential step as your understanding of the current business situation is foundational to the next tasks and you need to make sure that your understanding is correct and verified by the client.

When establishing the current situation, the first thing to clarify is whether there is already a data science function in the organization. If there is, the current situational analysis should be based on the following questions:

• What kind of models do they build and why?
• What technology do they use?
• What is their current perception of Pega Customer Decision Hub?
• Do they understand the data science capabilities of Pega Customer Decision Hub?
• Do they buy into the 1:1 customer engagement paradigm?
• What is the structure of the team?
• What kind of model governance is there?

If there is no data science capability, the investigation should be based on the following questions:

• What is their perception of data science?
• What is their current perception of Pega Customer Decision Hub?
• Do they understand Pega Customer Decision Hub data science capabilities?
• Do they buy into the 1:1 customer engagement paradigm?
• What are their aspirations in this area?

 

Going back to our common approach, let's continue with common step 2, which is to enable the client on relevant Pega best practices.

 

Pega has developed best practices in each of the organizational transformation areas based on deep product knowledge and years of delivery experience with clients across the globe. Best practice materials are covered in future topics, however, in terms of approach, it is important to consider adapting the delivery of these materials according to your stakeholder analysis and analysis of the current situation.

For this topic, the main cause for adaption will be the level of maturity of the current data science function in the organization.

 

For clients with a well-established data science function, it is important to consider the following:

 

• They need to understand clearly how Customer Decision Hub adaptive and predictive models work in great detail.
• They need to understand their responsibilities in terms of setup, model governance, maintenance, and optimization.
• They would need to be involved in an initial review of predictors and the quality of data going into them.
• They need to understand how the models in Customer Decision Hub fit in with the great work they are already doing. A key message here is that Customer Decision Hub adaptive models are fully automated, self-learning, and do not require pre-training. They are preconfigured for immediate use out of the box and learn fast. This means that they can meet the scale of demand required for a busy marketing function. The sophisticated models that the team is likely currently working on will remain as valuable as they are, and there are various ways their valuable insights can be used by models in Customer Decision Hub.
• They need clearly defined enablement paths for the above.

 

For clients with little or no data science capability, it is important to consider the following:

 

• They need to understand how Customer Decision Hub adaptive and predictive models work, but can do so in less detail,
• We need to present what machine learning in Pega applications can do for them in terms of creating valuable interactions with customers, and demonstrate how they are fully automated and self-learning. And, although maintenance and optimization are required, that there is a clear and easy interface to use to do this.
• They need to understand their responsibilities in terms of model governance, maintenance, and optimization.
• They would need to be involved in an initial review of predictors and the quality of data going into the predictors.
• We need to identify people in the organization who could be enabled to take on these responsibilities.
• They need clearly defined enablement paths for the above.

 

Common approach step 3 is defining a client-specific data science process for Customer Decision Hub. The first step is to assess the gap between best practice and the current situation. Then, take the best practice and design a future state that sets the client up for success and is feasible for them to implement within a reasonable timeframe. To be clear, this end state should be as close to best practice as possible, only diverging when it is absolutely not possible to adopt best practice.

 

It is essential to cover the following aspects:

• Setting up ADM for success.
• Stakeholders in the organization must understand the value of data science in the context of CDH.
• Resources must be identified to monitor and optimize the adaptive models in CDH – either data scientists or analytics people who understand and can review the quality of the data coming in (to start with, and eventually, a data scientist).
• Enablement paths for resources must be defined if required.
• The process for monitoring, optimizing, and reporting must be defined including identification of stakeholders that will receive information and the cadence for reporting.
• Configuration to support these processes is done in Prediction Studio.
• If external models need to be used alongside Pega applications, ensure this is set up and understood by stakeholders.
• Establish whether there is any historical data to be migrated.
• Ensure that model governance requirements are understood and can be answered to.

Work as collaboratively as possible with the client on this step to improve the likelihood of adoption. One example of how to do this is to work iteratively, incorporating their feedback. Another suggestion is to workshop elements of the future state.

If required, define several phases of change, such as by creating a change roadmap if the optimal state cannot be implemented from day one because of external restraints.

Document the future state as visually as possible to ensure understanding. Obviously, consider adapting documentation for different stakeholder groups, for example, by producing an executive summary.

Finally, present the solution – ideally someone in the business with whom you collaborated will present the solution, demonstrating full ownership and buy-in.

 

Finally, step 4, we define the implementation steps needed to transition from the current state to the future state. This ensures that the future state will actually materialize.

 

Once the future state has been agreed upon, make note of any tangible actions that need to be completed to get there. Ensure they are time-bound, have clear owners, and are agreed with the organization.

It is important to keep a risk and issue matrix to address anything that might impact the delivery of these actions. It is equally important to have regular checkpoints and an escalation route if matters fall behind.

You have reached the end of this video. What did it show you?

• The importance of Data Science Transformation in enhancing customer experiences using Pega Customer Decision Hub.
• The key stakeholders involved in the transformation process and their roles.
• The step-by-step approach to reviewing current data science capabilities, enabling best practices, defining client-specific processes, and implementing the transformation plan.