Enhancing Decision-Making through Data: Framing the Correct Inquiry

Nano Tools for Leaders?, a joint initiative of Wharton Executive Education and Wharton’s Center for Leadership and Change Management, offer rapid, impactful tools that can be grasped and implemented within 15 minutes. These tools possess the potential to significantly enhance your success, as well as the engagement and productivity of your team.

While most businesses recognize the value of data in facilitating informed decisions, the conventional approach often falters due to common missteps by decision-makers. Many leaders excessively rely on available data, which may not necessarily address the specific issue at hand, or delegate crucial decisions to data scientists lacking a profound grasp of the business challenge. Moreover, decision-makers frequently exhibit a predisposition, formulating a solution first and then seeking data to validate it.

Summarized from article by Stefano Puntoni and Bart De Langhe.

Action Steps

1. Get Clear on the Decision. Start by focusing on the decision you need to make. Stick to options that are within your control and relevant to your role. Most of us naturally lean toward familiar choices, so take the time to explore different perspectives by talking to your team or trusted colleagues. Your options should be realistic, impactful, and aligned with your goals—steer clear of anything too risky or costly.

Here’s the key: Don’t jump straight into the data. If you let available data drive the process, you risk asking the wrong questions. Instead, focus on the decision first, then figure out what data you actually need. While data exploration has its place, decision-driven analytics work better when you reverse the process: start with the decision, then find the data to support it.

2. Ask a Factual Question. When you need a prediction, ask a “factual” question. For example, manufacturers might ask, “When is this machine likely to break down?” so they can plan maintenance. Retailers might ask, “Which products are most likely to be returned?” With this info, they can adjust prices, take steps to reduce returns, or even stop selling certain products.

3. Ask a Counterfactual Question. When you need to compare what would happen with and without a certain action, ask a “counterfactual” question. These are more complex because they involve hypothetical scenarios. For example, during Barack Obama’s 2012 campaign, his team asked, “Who is most likely to be persuaded to vote for Obama if we contact them?” This saved resources by targeting swing voters who were most likely to change their minds, instead of wasting effort on voters who were already decided.

How This Plays Out Take HP’s “Instant Ink” subscription. If HP wants to reduce cancellations by offering incentives like discounts, they need to figure out whom to target.

If they ask a factual question, like “Who is most likely to cancel?” they’ll learn something useful, but it won’t help them decide who should get an incentive. Why? Because not everyone who’s likely to cancel will respond to the discount.

Instead, HP should ask a counterfactual question: “What effect will an incentive have on specific customers?” To answer this, they could run a randomized experiment—split customers into two groups, give incentives to one, and compare the results. This helps HP figure out not just if incentives work, but who responds best to them, saving resources and targeting the right people.

This shift from "What’s likely to happen?" to "What will make a difference?" is where real insights live.