Five tips for making wise, data-informed decisions

Five tips for making wise, data-informed decisions

Julian Geoffrey López


A person pointing to a laptop screen that shows data atop a map. The map focuses on a border between Tanzania and Kenya.

Development professionals want to use data to inform decisions, but often the available datasets don’t address the relevant questions or it’s unclear how the data was gathered. Sometimes using data to make decisions would even have a negative impact on colleagues and be strongly opposed by important stakeholders. What should you do?

Here are five tips for making data-informed decisions in these and other real-world scenarios.

1. Clearly articulate the information need

The first step is to articulate an information gap that, if resolved, would facilitate more effective decisions.

Ask yourself, “What evidence would be valuable and influential as I make this decision?” The foundation of a data-informed decision-making process is to formulate a question or problem statement to determine your objectives and better inform the data gathering, cleaning, and analysis.

For example, the human resources department at Google wanted to answer the following question: “Do managers make a difference in their team’s performance?” By asking this clearly defined question at the outset, the team was able to identify relevant data sources, such as employee surveys and performance reviews, to better inform hiring, training, and (when necessary) firing decisions about managers.

Too often in development, we find ourselves looking at available data first and then searching for a problem or decision the data could help us address. In contrast, starting with a clearly defined problem and question before collecting data helps to ensure that the data you collect will meet a real need.

2. Ensure data is representative of all groups

Employing a diversity, equity, and inclusion lens in every decision might be time-consuming, but it is necessary. If you lead a development organization or a large project, you’re likely receiving regular reports from staff with insights and recommendations. As a decision-maker, you have a responsibility to review the disaggregate data and ask how it was gathered and analyzed.

For example, imagine your organization conducts an annual staff survey to inform its strategic priorities. To understand which supervisors need help in boosting morale, it isn’t enough just to look at the overall results; you must also break down the data by team and make sure that you have an adequate survey response rate that represents most—if not all—employees.

If data collection seems burdensome and expensive, leaders should consider adopting the type of lean data approach developed by Acumen and Harvard Business School to tackle the complexities of data usage in decision-making. Never be afraid to iterate, gather more data, and analyze it again. If you’re not willing to take a step back after identifying a data gap, the evidence and conclusions will be flawed.

3. Identify your biases—and those embedded in data

In an increasingly digital world, leaders have access to more data than ever, but human bias can easily warp which data gets collected and how it’s analyzed.

Be wary not just of your own preconceived notions when interpreting data but of the human biases that have been built into machines.

Many companies have created algorithms that perpetuate biases. As Cathy O’Neil explains, “Algorithms are opinions embedded in code.... [They] repeat our past practices and automate the status quo.” Her book Weapons of Math Destruction is essential reading for development professionals who work with data.

She gives the example of a company that wants to use machine learning to define the characteristics of a successful job applicant and select applicants for interviews. Machine learning is driven by historical data. If this company has only hired men in the past, then the algorithm would use gender as a predictor of success and skew the candidate pool toward men.

Sometimes biases can be even more difficult to detect in results. For example, Amazon found that its system was unfairly downgrading graduates of two all-women’s colleges.

Researchers can also unconsciously introduce sampling bias during the data-collection stage. Leaders must be aware of groups that might not be included in the data instead of accepting the preliminary analysis at face value or jumping to premature conclusions.

4. Make data-informed decisions that align with your values

To make decisions, we must be data-informed and not data-driven. The former recognizes that there are intangible variables that influence decision-making. The latter assumes that data is 100% accurate and that the only factors that matter are the ones we can quantify.

We could gain access to all the data in the world, but at the end of the day, our values and experiences will drive our toughest decisions. A study found that “61% of U.S. business decision-makers believe that human insights should precede hard analytics when making decisions.”

To better recognize the elements influencing your decision, you can use value maps or matrices to prioritize your values and understand their significance relative to a decision.

Another approach to guide your decision-making process is to revisit your organization’s mission, values, and goals to identify how well they align with decisions. This is a particularly effective strategy when gathering feedback from team members and stakeholders about specific dilemmas. 

5. Be transparent about your decision-making process and explain how it was data-informed

Often, leaders must make controversial decisions that will have negative repercussions on their colleagues or business partners.

The worst way to communicate tough decisions is to be impersonal and evasive—for example, by sending an ambiguous mass e-mail to all employees.

Instead, work to achieve transparency and empathy before and after making data-informed decisions. This can reduce the negative impact on morale and make employees more likely to accept the outcome while guiding you toward better decisions.

Having an honest conversation with your colleagues about how the data was collected, discussing the known biases within it, and acknowledging that the data does not encompass all aspects of people’s experiences will build trust. In communicating the final decision, explain how the decision was made and who was involved in the deliberation. Lastly, create open channels of communication that will enable all stakeholders to have a dialogue with you regarding the decision, its impact, and the outlook for the future.