Digital transformation is a buzz phrase at the forefront of many business plans today. Data, of course, is at the heart of it. But who has access to that critical data for decision-making? If your data is controlled by a coterie of elite specialists, even though not many outside that group may show much interest, your organization is likely to be far less effective, argues Brett Hurt, CEO and co-founder of data.world, in this opinion piece.
Consider the last time your team used data to make a decision. Were stakeholders with relevant knowledge throughout the business involved in the process? Or did they learn about the decision after it was made, without their input? Were the people doing the data analysis also experts in the subject matter? If you’re a high-level decision maker, you’re probably completely removed from how data is collected, organized and analyzed. Any plan, however brilliant seeming, stands on loose pillars if it comes from a team that is long on data talent and short on domain knowledge. This scenario, unfortunately, is the norm in corporate America.
Here’s a second scenario. You uncover a business problem that data can help solve. The employees who deal with the challenge daily, who feel its pain acutely, get immediately looped in on the project and consulted throughout its progress. You make the data available with the context they need to understand it. They ask questions only people like them would think to ask. Your data scientists and analysts probe the data and come back with illuminating answers. Everyone speaks the same language with the help of a shared data dictionary. This inclusive approach to collaboration capitalizes on the amazing array of perspectives, skills and knowledge available at any sizeable company. And deep data collaboration increases the data literacy of your non-quants while improving the business knowledge of your data people. As a result, the next project begins on better footing, and its result is even better than the last.
If the first scenario is more familiar, you have a data elitism problem. Data elitism happens when decision-makers only focus on ways to make advanced users more productive with data. Put another way, data elitism is any practice that makes data less inclusive. It’s usually not intentional — it’s more of a default state.
Most businesses will never realize more than a fraction of data’s potential. They leave out too many of the people with valuable knowledge and skill. According to Qlik, only 33% of full-time employees are confident in their data literacy. Even more frightening, the study revealed that “55% of employees lack the education and resources to make sound decisions based on insight, and rather make decisions based on ‘gut feelings.'”
“Most businesses will never realize more than a fraction of data’s potential.”
And these problems ripple upwards. Because so much of a typical business’s workforce is unable to find data, ask clear questions about it, understand it, and share what they know about the problem, data analysts and data scientists get weighed down helping with low-impact tasks instead of the high-impact work they were hired for. They answer the same basic questions repeatedly. They play traffic cop, pointing people to data they can’t find themselves. And they use powerful tools to run simple analyses because most data tools that claim to be “self-service” come with extreme learning curves, complex technical jargon and bad interfaces that scare away common business users. In this context, data elitism becomes a serious limiting factor.
So what leads to the data elitism problem? Often it’s a combination of structural and personnel factors. Here are some examples:
- Unused data: People don’t know what data your company has and can’t find data when they need it. How can you use something if you don’t know it exists?
- Dark data: Too much data work and analysis takes place without transparency, audibility or accountability.
- Homogeneity: Project outcomes reflect the knowledge, abilities and biases of a small, homogeneous group. These advanced practitioners convey insights that seem cryptic to most of their colleagues.
- Puzzling barriers: Access controls and other restrictions persist without regular reconsideration and changes.
- Misguided data spending: Fancy tools can boost the productivity of data elites but they don’t lift the general data literacy baseline.
- Dominating HiPPOs: The Highest Paid Person’s Opinions are taken more seriously than the data; these individuals opt to do what they want because they know best.
- Undervaluing non-STEM contributions: How often are non-STEM employees closest to the business challenge at hand? How often are they dismissed from the larger conversation?
Collective Data Empowerment
Data elitism is a serious problem for businesses lagging on the transition to the digital age. Crossing this huge data divide can only be achieved by building a data-driven culture. This process begins with collective data empowerment, instituting measures that make everyone in your organization more productive with data.
Airbnb provides a great example of this holistic approach. Despite making data more accessible to employees and building remarkably sophisticated data tools, only 30% of employees were using the internal platform on a weekly basis in Q3 2016. Airbnb built its internal Data University to “empower every employee at Airbnb to make data informed decisions by providing data education that scales by role and team,” according to Jeff Feng, Product Lead, Data. The 30-course program educates people “in the context of Airbnb’s data,” and instructs users on the specific data tools used at Airbnb. Less than a year after the company benchmarked Weekly Active Users of its data platform at 30%, Airbnb reported 45% of its workforce uses the platform weekly.
Of course, Airbnb is a famously savvy modern tech company. Your company might be bigger, slower and less excited by radical change. To beat data elitism, start by taking stock of who uses data, what’s broken, and what’s going well. I recommend trying the following three practices, which we have documented in detail here.
- Data project ride-alongs: Choose a typical data project and check in with stakeholders at key stages to observe how their tools, objectives, needs, collaborators and challenges change. You will ask yourself questions like, “Whose perspective or skill is missing from this project?”
- Data project post-mortems: Squeeze more knowledge out of every data project by creating a simple process to apply lessons learned to the next project. Try to conduct post-mortems within a week of completing a project. Make sure to review project goals, give all participants a voice, highlight accomplishments, discuss improvement areas and lessons learned, and finish with outstanding action items.
- The Manifesto for Data Practices: Speed-up collective data empowerment by committing to the Manifesto for Data Practices, a set of four values and 12 principles that we believe describe the most effective, ethical and modern approach to data teamwork. I co-authored the Manifesto with dozens of leaders from diverse backgrounds including academia, business, journalism, open source and the public sector. More than 1,500 people from every flavor of company have signed so far.
Through a thoughtful combination of tools, practices and strategies you can take better advantage of your business’s available knowledge and talent. It starts by identifying the areas where your company’s data practices and technologies exclude more people than they invite.