What Are Three Things You Need to Do to Foster a Data-Driven Culture?

According to the worldwide bestseller Thinking, Fast and Slow by Nobel Prize winner Daniel Kahneman, people make decisions either intuitively or logically. Intuition leads to fast decisions; rational thinking leads to slow decisions. In organizations, it’s the other way around. Intuition leads to long decision-making processes; data- and fact-driven decisions lead to shorter processes.

In intuition-driven cultures, different people follow their unprovable hunches. Opinions collide, and in the end, the winner is the one who can tell the best, most fantastic story—or the one who is the highest paid. Data-driven organizations typically make decisions faster, with less debate and a higher probability of success.

However, adopting a data-driven culture is not easy; established behaviors must change, and the data necessary for decision-making is often unavailable and cannot be interpreted correctly.

Change Statements to Questions and Experiments

Experience, intuition, and opinionatedness are important qualities that help initiate discussions and decision-making processes. In intuition-driven organizations, people jump quickly from opinion to solution. Opinions are often expressed with an exclamation point and met with another exclamatory opinion.

In data-driven organizations, opinions end with a question mark and elicit the response: “Let’s try it out.” An opinion leads to a question, which leads to a hypothesis that is tested and validated on a small scale. The data and experiment results drive decisions made by those with experience weighing all aspects and consequences (not by automated AI).

It is not intuition that feeds the decision-making process but data from a targeted experiment

Change Randomly Storing Data to Purposefully Generating Insights

Organizations that want to adopt a data-driven culture often find they have no data, too much data, or nonmeaningful data. This is usually because data is stored without a plan. Data warehouses and data lakes become flea markets where you can find valuable things if you know your way around, but most items are useless and not worth the cost.

Data-driven organizations generate data specifically for a particular issue. They define and implement opinion-validating experiments to generate the exact data and insights needed. They use different technologies to store, process, analyze, and visualize data depending on the question, semantics, and syntax. This also helps with data protection because the data is collected for specific, explicit purposes (purpose limitation) and is limited to what is necessary for the reasons why it is processed (data minimization).

Amazon Web Services provides technologies to help build an end-to-end data strategy. (For more information on how to reinvent your organization with data, visit Amazon Web Services for Data .)

Change Storytelling to Data Literacy

In intuition-driven organizations, telling stories using endless PowerPoint slides is an essential skill that often ends up in dog and pony shows more reminiscent of circuses than decision-making meetings. Data and statistics are used—but only to support the “story.” These statistics are often taken out of context, and the forecasts, in particular, are statistically insignificant.

In many organizations employees understand totals and averages. But they don’t understand medians, standard deviations, percentiles, or cohorts. They don’t know how to properly generate, prepare, and especially visualize data. Some organizations introduce a newer, nicer, more colorful business intelligence tool almost every year but forget to train employees on basic data literacy, let alone how to use the tool.

In data-driven organizations, on the other hand, storytelling is deliberately limited in favor of data-based reasoning. For example, we at Amazon Web Services use documents instead of presentations as a basis for internal decision-making.

This way, everyone can be a decision-maker. Decisions are made faster and with the customer in mind.