Scaling Data Maturity With Successful and Proven Frameworks

Reading Time: 9 mins

Becoming a data-driven organization often requires a paradigm shift in organizational mindset and culture. It involves investments in upskilling, infrastructure, and technology.

During their data transformation journey, businesses sometimes need to prioritize their projects. On 14th Nov 2022, Ganes Kesari, the Chief Decision Scientist and co-founder of Gramener, sat down with Adel Nehme of DataFramed, a podcast for data and analytics leaders, to discuss the processes and frameworks that can help organizations become data-driven, gauge data maturity and garner company-wide support.

In the episode, Ganes explains how organizations can scale their data science maturity and build a practical data roadmap. He also explains what it takes in terms of skill set and people management to achieve data maturity and more.

Below are some key takeaways from the podcast.

What is a Data-Driven Organization?

In simple terms, any organization that can effectively and consistently utilize data for decision-making across all levels can be described as a data-driven organization.

Data-driven organizations consider data a strategic asset, not just a support function or enabler. In addition to strategic decisions, these organizations also utilize data for everyday decision-making.

Data is a way of life for everyone within a data-driven organization.

What is the Most Challenging Aspect of Building a Data-Driven Organization?

Building a data-driven organization takes work. A recent survey by New Vantage Partners found that only 26%, or roughly one in four participants, had successfully created a data-driven organization. Less than 20% have data culture ingrained into the organization.

A business needs to undergo the following four shifts to become a data-driven organization:

Skill Set

The team building the solutions has to be upskilled in data analytics and storytelling. The people using these solutions will also have to upgrade their skill set.

Toolset

This involves all the tech stack – the technology, solutions, and architecture. It also includes all the investment required to build the tech stack.

Process Set

Integrate the tech stack into existent business processes so that it seamlessly aligns with the goals and objectives of the organization as opposed to a supplement that must be accessed separately.

Mindset

This is the most important but challenging of the four shifts. If you are unable to overcome the organizational culture resistance, make employees comfortable, and instinctively turn to data when faced with a problem, no amount of training, building solutions, or even rewiring the processes will be enough.

How Should Data Leaders Approach this Transformation?

Sometimes, organizations make significant investments of millions of dollars and take up mega projects for up to one year. Unfortunately, the risk of failure in such cases is high.

The chances of success are much higher when we take this step-by-step, deploying the four levers discussed in the last section. It’s best to start with the data, the analytical capabilities and tools, and the people we have.

We can build a roadmap to solve simple, descriptive, and smaller problems using our current capabilities. Once they are realized, we can project the benefits to the rest of the organization to showcase our achievements within a few weeks or months.

This is how we build momentum for future success.

How Do Data Maturity Frameworks Help Organizations Prioritize their Data Journeys?

Facing the prospect of a long-term data strategy, clients sometimes ask, do we have to go through all of this? Why can’t we just get started with a project tomorrow?

Data maturity reflects an organization’s readiness and capability to embrace data for decision-making. It helps you with your self-assessment and self-discovery. Before we can chart the path to where we want to go, we need to find out where we are.

At Gramener, we have a data maturity framework that has worked wonders for our clients. It comprises five dimensions:

• Vision – Does the organization have a short-term and long-term vision for data?
• Planning – Can the organization translate its vision into a roadmap, project set, and build capabilities?
• Execution – How well can the plan be executed? What kind of processes and tools do you have?
• Value realization – How are you able to improve adoption and measure RoI?
• Culture – Is there resistance to data within your organization? How ready are people to embrace a technology like this?

Some organizations are great at vision and planning. Their executive management is wholly aligned. However, they may need better execution or value realization.

Similarly, other organizations have a very strong execution muscle – good tools, good people, and a great data science team. However, they lack the vision to pick the right projects.

A company needs the right mix of all five dimensions, without which its investments could lead to failure. A data maturity assessment tells you at what level you are for each of these dimensions and where you need to focus next.

Data maturity improvement takes time. It is not something that can happen overnight. It is a journey that involves working on all these five dimensions.

What are the Different Levels of Data Maturity? What are the Investments for Success?

An excellent place to start would be the five levels of data maturity that Gartner has extensively published.

• Organizations that are at level 1 use data opportunistically. Businesses like these turn to data only when they need help to overcome a significant hurdle through conventional means.
• At level 2, pockets within the organization start using data. This may include a few teams. It may also include a business or technology leader who decides to invest in it.
• At level 3, the adoption is broader than at level 2, but data still needs to be used strategically within the organization.
• At level 4, almost the entire organization uses data to realize business RoI.
• At level 5, the organization views data as a strategic asset central to business strategy.
• At every level of data maturity, a business must constantly improve its skill set, tool set, process set, and mindset.

How to Measure Organizational Data Maturity?

We should start with the self-assessment. Data leaders can use the right frameworks and toolsets to do this internally. Alternatively, they can hire an external partner.

To measure organizational data maturity, businesses should:

• Reflect on the organizational strategy and the role of data in the current state of the organization
• Talk to the business and technology teams and run surveys to understand their priorities. Where do they see the gaps in the data practices? How much business value does the data create for them?
• Avoid bias and blind spots by inspecting the assets. Data maturity surveys often see technology teams score higher than business teams.

While the business teams report that the availability of good quality data could be better, the technology teams need better visibility on the business priorities, which are constantly changing.

Inspecting the assets allows you to see which projects were run earlier, their value addition, and their shortcomings.

Combining these three elements can help business leaders gain an invaluable understanding of the data maturity level of their organization. It can help them assess the gaps that need to be bridged to realize their business goals and objectives.

What are the Steps to Becoming a Data-Driven Organization?

The first step in this five-stage journey is a business strategy or painting the vision for data. The data strategy vision will be governed and hence, your business strategy must be align with it. What are your business priorities in the short-term and long term? How do you see your current year shaping up? What is your 5-year plan?

The second step involves charting an execution roadmap to translate this vision. This includes identifying strategic initiatives to help achieve the targets for the upcoming quarter or year.

For example, suppose an organization is looking to grow its revenues and increase the top line in the current year. In that case, it may want to improve its customer relationship management or explore newer markets.

Using data strategically can help drive these business initiatives. It can help gain market intelligence to facilitate the choice of ideal markets for launch and expansion. It can also lead to topline growth by identifying factors that help expand the share of customer wallets.

Data can also help businesses answer the following questions:

• What is my current share of the customers’ wallets?
• What are the things that I am doing right?
• Where is my room for improvement?
• What will make my customers switch from my competitors to me?

These business priorities will drive the data strategy roadmap.

The third step comprises strategic initiative prioritization. We build our capabilities based on the projects that we want to execute. For instance, a company may decide to undertake diagnostic and descriptive projects for the upcoming two quarters, which will not require hiring AI experts.

This step is also crucial in prioritizing projects. Building capabilities should always follow project prioritization. Sometimes, businesses hire 5-10 PhDs in data science before identifying projects they want to execute, similar to putting the cart before the horse.

The fourth step, once you identify which capabilities to build, involves building the right tools. It also involves identifying the business areas that require rewiring and planning for the process set integration.

The fifth step comprises the mindset shift.

Following these five steps is not a linear process. They have to be repeated in multiple cycles that last between two to three months, after which they are revisited and expanded again and again.

Best Practices to Build a Data Science Roadmap

Since most firms focus on the short-term and dedicate limited resources, 1-to-2-year roadmaps are a common requirement from clients. Some of our clients ask for 5-year roadmaps. Most projects have a timeline of 12 to 18 months, while aspirational initiatives last for two years or more.

The ideal way to build a data science roadmap is to start with the business priorities, balance impact, and feasibility, and balance short-term and long-term initiatives. We can visualize this using a 2X2 matrix.

Feasibility refers to the availability of data, tools, and people to execute a project. Projects with high feasibility and impact should ideally be at the top of the execution priority list.

The quick turnaround time frequently tempts clients to execute projects with low impact but high feasibility. The key to prudent decision-making here is to balance impact with feasibility. Is the solution going to offer a key breakthrough? If not, there is no point in executing it.

When building a 12-to-18-month roadmap, you should have a good mix of short-term and long-term initiatives. Strategic initiatives that require more investment but deliver long-term benefits should feature in equal measure with quick wins.

Where Do the Biggest Business Benefits Come From?

Business leaders often feel that the biggest benefits come from predictive analytics, where they should focus their investments. While investments in predictive analytics can deliver rich dividends to a business, especially in the long run, simpler analytical tools with diagnostic and descriptive capabilities can provide quick and impactful wins.

For example, we helped one of our clients, a manufacturing firm, discover the reasons behind the failures with the batches produced, delivering a high-impact solution using diagnostic analytics.

Similarly, projects that identify the two or three major drivers of yield would be immensely beneficial in the short term to any manufacturing organization.

To be sure, predictive analytics engagement has undeniable advantages. For instance, manufacturers rely on multiple factors to produce the golden batch in the pharma industry. Being able to predict the machine parameter setting can help identify the factors that drive the optimal yield.

Within predictive analytics, a category of use cases, such as a simple customer churn model, doesn’t need to be embedded in a business process or software. It just shows the group of customers more likely to churn.

One of our clients, a telecom firm, was trying to identify which customers are more likely to churn. Some simpler models, like Decision Trees, improved the conventional or manual methods by up to 35%. More advanced techniques, such as Deep Learning, improved the accuracy by up to 60% – a difference of up to 25%!

However, the more advanced techniques brought with them their own set of challenges. First, there was the additional engineering cost. Second, solutions like the Decision Tree offered reasons why corrective actions had to be taken to prevent customers from leaving.

On the contrary, Deep Learning methods share a set of customer names that are likely to churn without any explanation. Marketing teams were averse to acting against customers without reason. In some cases, they even cited evidence that the said customers were heavy users of their products and services.

With advanced techniques, user-friendliness becomes a challenge. Even with predictive analytics, it is prudent to start with simpler tools.

Finally, How to Ensure the Right Execution of a Roadmap?

Sometimes, the projects that are greenlit for execution differ from the roadmap. This may occur because of dynamic business priorities, unforeseen crises, or a change in leadership perception.

While revisiting the roadmap regularly is recommended, it should always be done keeping short-term and long-term business impact and feasibility in mind.

Another familiar mistake chief data officers make is while allocating projects across functions. If 4-5 projects could distribute across 4-5 functions, CDOs sometimes allocate one project to each function to democratize the process.

Unfortunately, for each function to realize RoI on its data projects, it usually needs to build a minimum portfolio of 2-3 projects in adjacent areas with high synergy. Otherwise, these functions will not be able to demonstrate any sizeable benefits from the projects.

The upper echelons of management are in a unique position to decide upon and reflect the benefits of a roadmap to an organization. In addition to signing the cheques, they should also review and greenlight the initiatives.

They can also suggest a viable alternative if an initiative needs to be scrapped or replaced.

A data & analytics steering committee is an excellent vehicle to include the top management in the roadmap. This committee can comprise both technology and business leaders.

The steering committee can meet periodically to ensure that the roadmap aligns with the organization’s business vision. They can also review and pick the data projects that will help the roadmap deliver RoI.