For many organizations, data-driven solutions rarely drive decision-making. The explosion of data can usher in a new era of fact-based innovation in organizations, backing up new ideas with hard evidence. Although it is simple to describe how data can be integrated into the decision-making process, it is much harder to make it regular or automated for employees — a culture shift in thinking that presents a daunting challenge.
We’ve distilled ten essential steps to creating and sustaining a data-centric culture:
1. Start at the top
Organizations with exemplary data-driven cultures often have executives who expect decisions to be firmly based on data. Whether it’s sifting through the results of trials or starting each meeting by examining data-backed proposals, senior executives need to lead with evidence-backed actions. These practices trickle downwards because employees who want to be taken seriously must communicate with managers on their terms and in their language. The example of a few people at the top can lead to significant changes in company-wide norms.
2. Pick metrics wisely
Executives can influence behavior by choosing what to measure and what they want their employees to measure. Suppose a company can profit by predicting the price movements of its competitors. Then teams should constantly make precise predictions about the magnitude and direction of this movement with data-driven solutions. More importantly, tracking the quality of those predictions will lead to steady improvements over time.
3. Enable data scientists
Data scientists are often isolated within organizations, resulting in a lack of knowledge exchange between them and business leaders. Analysts cannot survive or create value if they are separate from the rest of the organization. Those who have successfully tackled this challenge generally do so in two ways.
The first strategy is to remove barriers between business and data scientists. Some companies task data employees with scaling their proofs-of-concept at the frontlines before they can return to their traditional roles. Others create new roles across functional areas and lines of business to increase analytical complexity.
In addition to bringing data science closer to the business, leading companies rely on the second tactic of pulling the business toward data science, primarily by encouraging employees to become competent at code and conceptually proficient in quantitative topics. Senior leaders do not need to be reincarnated as machine learning engineers, but leaders of data-centric organizations cannot ignore the language of data.
4. Streamline data access
One of the most common complaints is that employees from different areas often struggle to access even the most basic data. Oddly, this situation persists despite many efforts to democratize access to data-driven solutions within organizations. Without adequate information, analysts don’t do much analysis, and a data-driven culture is unlikely to take root, let alone thrive. Instead of comprehensive but slow endeavors to reorganize all data, sometimes it’s better to focus on just some data, to begin with, and build on from there.
5. Measure uncertainty
Although absolute certainty is impossible, asking teams to provide a corresponding level of confidence in their work can help clarify and quantify the unknown variables. Decision makers then have to deal with potential sources of uncertainty directly and understand whether the data is reliable. Analysts must critically assess uncertainty and gain a deeper understanding of their models. Finally, the emphasis on understanding uncertainty motivates companies to experiment.
6. Keep prototypes simple
There are often many more promising ideas than realistic solutions. The differences only become apparent when a company attempts to turn a proof-of-concept into production. Data-driven solutions help to develop industrial-grade proofs of individual straightforward concepts. Then, with more data, each component can be gradually built up independently with larger datasets or outlandish models to determine what works best.
7. Complement training with experience
Many companies invest in all-at-once training days for employees to quickly forget what they’ve learned. New skills lose their importance if they don’t apply them right away. While basic skills like programming should be part of basic training, it is more effective to train employees before they need specialized analytical concepts and tools. Directing teams to hone their skills on proof of concepts will forge deep-rooted knowledge, and once-unfamiliar concepts behind data-driven solutions will become commonplace.
8. Help employees help themselves
It’s easy to forget the potential role of data fluency in employee satisfaction. Empowering employees to play with data can significantly improve their work, as they can automate tedious, monotonous tasks. When teams interpret the concepts behind data-driven solutions, they can improve their work by saving time, avoiding rework, or accessing often-needed information. When it comes to internal upgrades, data-fluent employees can participate in creating a working solution instead of issuing a wishlist to the IT department.
9. Standardize metrics for clarity
Many data-dependent companies share the problem of teams picking their preferred bespoke data metrics and programming languages. Across a company, this can lead to abundant issues. Businesses can waste countless hours trying to reconcile subtly different versions of a metric that should be common. Inconsistencies in the way modelers work can also take a toll. When coding standards and languages differ across an organization, analytics talent requires retraining at every step, making it difficult to spread. Sharing ideas internally can also be cumbersome if translations are always needed. Businesses should choose standardized metrics and programming languages to benefit in the long term.
10. Document analytical choices
Analytical data problems rarely have a single solution, and data scientists must evaluate different tradeoffs to pick the best approach. Asking the team how they approached the problem, what alternatives they considered, what tradeoffs they understood, and why they chose one method over the other leads to a deeper understanding of the solution. Analyzing choices helps to consider broader alternatives, reconsider potential assumptions, and create a robust evidence-based foundation for future teams.
Conclusion
Data can provide evidence to support hypotheses and give managers the confidence to enter new areas and processes instead of stumbling in the dark. However, data-driven solutions alone are not enough, and companies must develop a culture where these solutions can thrive. Leaders can lead this change by example, practicing new habits, and creating expectations about evidence-based decisions.
Data science is more than methods, machine learning, artificial intelligence, and related tools. To adapt your organization to this emerging field, you must make the necessary changes to your company’s existing ecosystem and prepare for a data-driven culture. Your team’s aligned goals, mindset, and data science approach will undoubtedly open up opportunities for your organization to reach new heights like never before.
At Itirra, we provide companies with customized solutions and opportunities to explore potential improvements to support their business goals. If you want to find out more or to discuss how we can help your business, contact us or schedule a meeting with me.