Sign up for our newsletter and get the latest big data news and analysis.

How to Choose the Right Organizational Model for Data Science and Analytics

In this special guest feature, Martijn Theuwissen, a co-founder at DataCamp, highlights how the most competitive companies prioritize developing data fluency across their workforce to improve outcomes. Organizations that aren’t able to effectively make use of their available data today are already behind the curve. DataCamp is an online interactive learning platform with over 4 million learners that focuses on building the best learning experience for data science and analytics. It offers 1000+ hours of learning through courses, projects, skill assessments and mobile practice. If Martijn is not tinkering on new learning interfaces, you can probably find him reading random stuff on Quora.

Data Fluency for Organizations

Gartner estimates that by 2020, 80% of organizations will begin to implement corrective measures to improve their data science and analytics capabilities. Organizations that aren’t able to effectively make use of their available data today are already behind the curve. The most competitive companies prioritize developing data fluency across their workforce to improve outcomes. 

What It Takes to Be Data Fluent

So, what is data fluency? Data fluency is the ability to understand data, communicate insights from that data, and ultimately to make more informed decisions. It’s about empowering employees with the skills to drive better business insights, faster. The goal is to democratize data science and build data skills at appropriate depths across the organization.

But just as with learning a language, data fluency is on a spectrum of proficiency. Each employee’s proficiency and comfort with data should reflect the needs of their job, which means data fluency should scale in relation to the complexity of the business problems at hand. Organizations must nurture a learning environment where employees are encouraged to grow their data skills according to business needs.

What It Looks Like to Be Data Fluent

Fostering a culture of data fluency is crucial for any organization that prioritizes faster, smarter, and scalable work. Even a basic understanding of data tools and resources can greatly improve the quality of interaction among individuals and teams, along with the efficiency of the organization as a whole. Speaking the common language of data allows a deeper mutual understanding of problems, methodology, and results.

In practice, data fluency will mean different things to different teams, but the outcome will always be better data-driven decisions. For marketers, it may be gaining insights to plan, execute, and measure a successful campaign. For finance teams, it may be using predictive analytics to more accurately forecast demand. For HR, it may be determining best practices for recruiting and employee engagement. And for executives, it may be creating a vision for an analytics strategy and building a data foundation, including establishing the company’s data science and analytics model.

Optimizing Data Fluency Across a Company

Structuring data science and analytics within a company is really about optimization of talent. Within departments, acquiring data skills directly drives insights and business decisions. But executives who oversee those functions also have the power to organize them for greater efficiency and impact. Let’s take a look at the different ways companies can organize their data function.

Organizational Models for Data Science and Analytics

Among our customers and more broadly, we see companies using three main models to organize their data teams: centralized, embedded, and hybrid. 

The Centralized Model

The first model is a centralized data science and analytics team that fields requests from other departments, commonly set up as a Center of Excellence. In this model, the Chief Data Scientist or Chief Data Officer oversees the entire data function and is able to prioritize data needs independently of other functions. The centralized data team is essentially the gatekeeper of data.

This model is widely used but can be problematic because it creates a silo for data tools, skills, and responsibility. Employees outside of the data team are not encouraged to develop data fluency for themselves because they view it as someone else’s responsibility. It may also mean that the data team is fully at the mercy of other teams’ requests, and unable to have full ownership to establish its own strategic vision. Likewise, requestors are at the mercy of resources available on the data team. Either specialists are available or they aren’t, and requestors typically do not have visibility into the progress being made. There may also be a lack of communication and mutual understanding along the way, as the data team and the requestors may have different ideas on prioritization and delivery. 

The Embedded Model

The second is the embedded, or decentralized, model, where data professionals are embedded in functional teams. You might have a data scientist or analyst on the marketing team in charge of marketing analytics, another on the sales team supporting sales targets, another on the engineering team supporting data infrastructure, and another on the finance team in charge of financial modeling. The data professionals report directly to the heads of each department function and are in the loop with team needs, so they typically have a better understanding of the needs at hand than in the centralized model.

But the embedded model can be problematic. Department heads may not be data fluent, meaning they may not be able to provide proper guidance and support for their data scientists. There may also be a lack of analytics standardization across the company resulting in decentralized reporting. 

The Hybrid Model

The third is the hybrid model, where there is a central data team and data professionals are also embedded in functional teams. As in the embedded model, they report to department heads, but they also have a dotted line to the Chief Data Scientist at the company. This allows for the Chief Data Scientist to bridge the gap between data needs and functional expertise, acting as both a visionary and a technical lead.

The hybrid model encourages more cross-functional collaboration and can enable a strong sense of purpose for each data professional. While it’s very powerful, it requires careful organizational structuring and is generally favored by companies whose data fluency is more mature.

Adapting to Evolving Data Needs

Business leaders should consider employee needs when choosing their data science and analytics organizational model. It’s no secret that skilled data experts are in short supply, which means they have a lot of options. For employers, this means employee retention can be a problem and they must ensure their data fluent employees are sufficiently engaged.

Business leaders should also consider their company’s current and future needs when choosing their organizational model. An incremental approach would be wise. Early-stage businesses can take a lean approach with a centralized data function supporting the whole company to start. Data scientists and analysts command high salaries (commonly in the six figures) and many early-stage businesses may not be able to afford to staff one on each team. Then, as analytics needs and capabilities scale, the company may pivot to an embedded model. As the company matures, it may choose to adopt a hybrid model to boost operational speed and extend capabilities. 
McKinsey asserts that the most essential part of implementing change successfully is establishing a strong culture that can keep up. It all comes back to fostering data fluency across the entire company. No matter which data science and analytics organizational model your company has today or chooses to adopt in the future, getting everyone to speak the language of data is what’s most important.

Sign up for the free insideBIGDATA newsletter.

Leave a Comment

*

Resource Links: