Modern Organisations are swimming in data. However, despite investing heavily in data collection, technologies, and talent, many companies struggle to create a strong, data-driven culture. The biggest obstacles to achieving this goal are not technical, but cultural. In this article, we will explore 10 essential strategies for creating and sustaining a culture with data at its core, drawing on real-world examples and practical advice to help your organisation become truly data-driven.
Leading by Example: The Role of Top Management
A data-driven culture starts at the very top of an organization. Companies with strong data-driven cultures typically have top managers who set the expectation that decisions must be anchored in data. They lead by example, demonstrating that data-based decision-making is the norm, not the exception.
At one retail bank, C-suite leaders collectively analyse evidence from controlled market trials to make product launch decisions. Similarly, senior executives at a leading tech firm dedicate time at the beginning of meetings to review detailed summaries of proposals and their supporting facts, enabling them to take evidence-based actions.
When top managers prioritise data-driven decision-making, this mindset cascades down through the organisation, as employees recognise that they need to communicate with senior leaders using data and evidence.
Selecting Metrics Strategically
Leaders can significantly influence behaviour by carefully choosing what to measure and which metrics they expect employees to use. By aligning metrics with desired outcomes, companies can drive progress towards their goals. For instance, if a company aims to profit by anticipating competitors' price moves, it should establish a metric for predictive accuracy over time.
Teams should then continuously make explicit predictions about the magnitude and direction of these moves and track the quality of their predictions, which will inevitably improve with practice. In another example, a leading telco operator seeking to provide key customers with the best possible user experience realized that it lacked detailed data on individual customers' service quality.
By creating metrics focused on customers' experiences, the operator could quantitatively analyse the consumer impact of network upgrades, empowering data-driven decision-making.
Integrating Data Science into the Business
Data scientists often work in isolation within a company, resulting in a disconnect between their work and the needs of the business. To bridge this gap, successful organizations employ two key tactics.
First, they create porous boundaries between data scientists and the rest of the business. One leading global insurer rotates staff out of centres of excellence and into line roles, where they scale up proofs of concept before potentially returning to the centre. A global commodities trading firm has designed new roles in various functional areas and lines of business to augment analytical sophistication, with dotted-line relationships to centres of excellence.
The second tactic involves pulling the business towards data science by requiring employees to be code-literate and conceptually fluent in quantitative topics. While senior leaders don't need to become machine-learning engineers, they cannot remain ignorant of the language of data in data-centric organizations.
Addressing Data Access Issues
One of the most common complaints in organizations is the difficulty employees face in obtaining even basic data. This issue persists despite efforts to democratize data access within corporations. Without access to information, analysts cannot conduct meaningful analyses, and a data-driven culture cannot take root.
To resolve this gridlock, leading companies provide widespread access to a select few key metrics at a time, instead of trying to reorganise all their data in one go. For instance, a prominent global bank established a standard data layer for its marketing department, prioritising the most pertinent metrics to predict loan refinancing requirements.
By prioritizing the metrics on the C-suite agenda and demanding that other numbers eventually tie back to this data source, companies can dramatically encourage data utilization.
Embracing Uncertainty
While everyone accepts that absolute certainty is impossible, most managers still ask their teams for answers without a corresponding measure of confidence. By requiring teams to be explicit and quantitative about their levels of uncertainty, organizations can reap three powerful benefits.
First, decision-makers are forced to grapple directly with potential sources of uncertainty, such as data reliability, sample size, and emerging competitive dynamics.
Second, analysts gain a deeper understanding of their models when they must rigorously evaluate uncertainty.
Finally, an emphasis on understanding uncertainty pushes organizations to run experiments. As one retailer's chief merchant noted, "At most places, 'test and learn' really means 'tinker and hope.'" In contrast, at his firm, quantitative analysts and category managers conduct statistically rigorous, controlled trials before implementing widespread changes.
Designing Simple and Robust Proofs of Concept
In analytics, promising ideas far outnumber practical ones. Often, the difference becomes apparent only when firms attempt to put proofs of concept into production.
To avoid discarding good ideas due to implementation challenges, companies should engineer proofs of concept with viability in production as a core component. One effective approach is to start by building something industrial-grade but simple, and then gradually increase sophistication.
A data products company implemented new risk models on a large, distributed computing system by starting with an extremely basic, end-to-end process using a small dataset. Once the foundation was in place, the firm could independently improve each component, such as increasing data volumes, introducing more complex models, and enhancing runtime performance.
Providing Just-in-Time Training
Many companies invest in "big bang" training efforts, only for employees to quickly forget what they've learned if they don't apply it immediately. Whilst basic skills like coding should be part of fundamental training, it is more effective to train staff in specialised analytical concepts and tools just before they are needed, such as for a proof of concept.
One retailer waited until shortly before a first market trial to train its support analysts in the finer points of experimental design. As a result, the knowledge stuck, and once-foreign concepts like statistical confidence became part of the analysts' everyday vocabulary.
Empowering Employees with Data
Data fluency can play a significant role in making employees happier by empowering them to automate mundane tasks and streamline their work.
When the idea of learning new data skills is presented in the abstract, few employees will be motivated enough to persist and revamp their work. However, if the immediate goals directly benefit them by saving time, avoiding rework, or fetching frequently-needed information, a chore becomes a choice.
For example, the analytics team at a leading insurer taught itself the fundamentals of cloud computing to experiment with new models on large datasets without waiting for the IT department to catch up. This experience proved invaluable when IT finally remade the firm's technical infrastructure, as the team could demonstrate a working solution rather than merely describe requirements.
Balancing Flexibility and Consistency
Companies that rely on data often harbor different "data tribes," each with its own preferred information sources, bespoke metrics, and favourite programming languages. While this diversity may seem beneficial, it can lead to countless wasted hours trying to reconcile slightly different versions of metrics that should be universal. Inconsistencies in modelling practices can also hinder the circulation of analytical talent and the sharing of ideas internally.
To address these issues, companies should establish canonical metrics and programming languages. One leading global bank insisted that its new hires in investment banking and asset management know how to code in Python, promoting consistency across the organization.
Fostering a Culture of Explanation
For most analytical problems, there is rarely a single correct approach. Data scientists must make choices with different trade-offs. To encourage deeper understanding and consideration of alternatives, organisations should make a habit of asking teams how they approached a problem, what alternatives they considered, the trade-offs they perceived, and why they chose one approach over another.
This practice often prompts teams to consider a wider set of alternatives or rethink fundamental assumptions. In one global financial services company, an initial assumption that a conventional machine-learning model for fraud detection couldn't run quickly enough for production use was challenged. The team realized that with a few simple tweaks, the model could become incredibly fast. Upon implementation, the company achieved remarkable improvements in accurately identifying fraud.
In Summary
Creating a data-driven culture is essential for organizations seeking to harness the power of data to drive innovation, improve decision-making, and gain a competitive advantage. While the journey to becoming data-driven can be challenging, the 10 strategies outlined in this article provide a roadmap for success.
By leading by example, selecting metrics strategically, integrating data science into the business, addressing data access issues, embracing uncertainty, designing simple and robust proofs of concept, providing just-in-time training, empowering employees with data, balancing flexibility and consistency, and fostering a culture of explanation, organizations can cultivate a mindset that embraces data as a fundamental driver of success.
Embarking on this journey requires a significant shift in thinking and a willingness to challenge established habits and processes. However, the rewards are substantial. Companies that successfully create a data-driven culture are better equipped to make informed decisions, respond to market changes, and identify new opportunities for growth. They are also better positioned to attract and retain top talent, as data-savvy employees increasingly seek out organizations that value and prioritize data-driven decision-making.
To begin your organization's journey towards a data-driven culture, start by assessing your current state. Identify areas where data is already being used effectively, as well as gaps and opportunities for improvement. Engage senior leaders in the process, securing their support and commitment to leading by example. Develop a roadmap that prioritizes initiatives based on their potential impact and feasibility, and establish clear metrics to track progress.
As you implement these strategies, remember that creating a data-driven culture is an ongoing process, not a one-time event. It requires continuous reinforcement, communication, and adaptation as the organization evolves and new challenges emerge. By remaining committed to the principles outlined in this article and empowering your employees to embrace data-driven decision-making, your organization can unlock the full potential of its data and thrive in an increasingly competitive landscape.