Digitization Maturity

Picture Credit: Shahadat Rahman on Unsplash.com

Digitization is more than a buzzword in the analytics' business vocabulary. Maybe not with as much bling, pomp, and mystery as AI, Machine Learning, or Deep Learning. Automation, RPA, bots, hyper-automation, digital-twin are some of the newer terms that have become part of the business executive's vocabulary and wish list in the last few years. The level of sophistication is increasing with many of these newer technologies. However, there are many practitioners just recycling the terminology to describe the 'same' work and efforts regardless of the sophistication. I tend to think of digitization as a catch-all term for all of these efforts. In this post, I'll touch on digitization maturity with business processes and the interplay with analytics maturity.

First of all, digitization of business processes is a journey. It's more like an ultra-marathon where the finish line seems to get farther and farther away the more progress is being made. It sounds paradoxical, however, this is a good sign 😄. This means the organization/team/leadership is stretching the scope of the digitization possibility and/or increasing the level of maturity. And along with this, comes the possibility of creating higher potential value and benefits reaped from the efforts. 

Analytics Maturity

A widely shown graphic on analytics maturity is the below Gartner Analytics Maturity Model. I have personally used this in documentation (slides) and in meetings. With a Google search, it is seen that this particular graphic and other variations of this widely exist. 

This provides a really good starting point. This is generally true for the level of capability that exists in an analytics team/org. It makes it simple and easy to understand from the point of 'We need to learn to walk before we run, we need to learn to run before we jump, etc, etc. ". However, there are some things implied here that are not correct when viewed in a more nuanced way. 

• Value creation and the type of analytics done to solve a business problem are not linearly connected/correlated. Some descriptive analytics initiatives can lead to larger value creation than a deep learning model run on the cloud :)
• In most cases, these capabilities are built on top of each other. However, individuals, teams, and organizations can jump levels. In my personal case, I did the prescriptive analytics first as I started out in my professional career - my main role was to wrote optimization models in GAMS that ran on the backend of various applications. This was mainly driven by my educational background and the point in my career i.e. individual contributor.  
• Achieving these different capabilities 'at scale' is NOT equally spaced apart i.e. it may take a team much longer to from step 2 to step 3 compared to going from step 1 to step 2. This is dependent on a lot of factors, the main one being the needed level of talent and the availability thereof of resources($) to get this talent. 
• The management and governance of these different types of analytics are not the same. 

A clear counterpoint to the Gartner model is the below list of steps done in a data science project as described by Hadley Wickham in his book. 

Visualize, Communicate = Descriptive Analytics

Model = Predictive/Prescriptive Analytics

And, these things happen in a circle?!?!? Such is the nature of good and sophisticated analytics work.

Source: R for Data Science book by Hadley Wickham

Digitization Maturity

Now, let's get to digitization maturity. The first part of digitization involves having data leading to analytics. This is a necessary step for the growth and state of analytics maturity in an organization. This is fully represented in the data science/analytics project steps in the figure above. 

The second part of the digitization efforts in an organization is converting the analytical output into decisions, actions, and eventually, outcomes for the organization. 

Just like analytics' maturity, there is a decision maturity aspect to how the analytics output can be converted into a decision/action. Almost all organizations will start with humans reviewing the analytics' output and making the decision. The extension of the digitization efforts will be the automation of these decisions followed by the automation of the next decision and the next action and on and on... This is the almost endless digitization journey of an organization. 

A recent article that highlights the need for 'decision impact' metrics along with model/analytics metrics gives a good summary of how to approach decision making. These concepts can be extended and applied on when to automate the decision making i.e. transfer the decision-making from humans to the digital process! Digitization of these decisions/actions will force the org to solve some hard problems. 

• Digitized decisions will have to be based on clear thresholds that maximize positive outcomes and minimize negative outcomes
• This will force or at the very least, pave the way to build some closed-loop systems to control the business processes. Closed-loop systems are great. 
• From the perspective of view of analytics/AI/ML/software development, generating this feedback loop is quite invaluable. 
• Achieving a closed-loop system is hard. It takes time, effort, and fighting the good fight (there will be nay-sayers in the org who just want the team to move on to building the next thing). 

There are many arguments made why a Human in the Loop (HITL) is necessary. This is a fallacy! In most industrial and business processes, this only represents an opportunity to grow the maturity of digitization. There are use cases in specific fields such as law enforcement, medical/pharma cases where HITL will always be needed.

A very good example to illustrate the expansion of the scope of digitization is cars and driving. A cruise control system in vehicles is a closed-loop system. Extending and extending the digitization spectrum has led to self-driving vehicles.

Digitization Maturity demands both analytics maturity and decision maturity. Teams and organizations have to focus, prioritize and execute on both these dimensions. 

The above framework for maturity/growth also sets up the roles and responsibilities within an organization nicely. 

• The analytics/software/data teams own the growth in the Analytics maturity dimension. 
• The responsibility to grow in the Decision maturity dimension should be owned by the 'business' or process owner/teams who sit within the functional organization. 
• This leads to shared goals and ownership i.e. skin in the game!
• It also has the potential of changing the supplier-customer (analytics team-business team) relationship into a true partnership!