Research: How Curveball Questions Can Surface the Insight You’re Looking For

We’re all taught to ask tough questions. Interviewers are coached to use open-ended prompts to get beyond rehearsed answers. Managers press for specifics when employees speak in generalities. Journalists and negotiators follow up when someone evades the point. The aim is the same: to uncover information the other person may be reluctant to reveal.

Yet questions that sound tough often fail to produce candid answers. Experienced respondents can anticipate them, prepare polished talking points, and steer the conversation toward what they want to discuss. They may answer at length while revealing very little. A question can be challenging without being revealing.

Our research, recently published in Strategic Management Journal, points to a different approach: asking a question the respondent could not have anticipated, but can still answer. We call it a curveball question. By disrupting the respondent’s script, a curveball makes it harder to rely on a stock response—and more likely that the answer will reveal how the person actually thinks.

Our research examined curveball questions in the context of quarterly earnings calls, in which executives first deliver prepared remarks and then respond to questions from financial analysts. Yet the logic underpinning the utility of curveball questions can be readily extended to such diverse contexts as board meetings, venture capital pitches, job interviews, and corporate strategy reviews.

The Case for Curveball Questions

We analyzed 126,910 quarterly earnings-call transcripts involving 5,609 U.S.-listed firms between 2011 and 2021. Executives begin these calls with a prepared presentation and then answer questions in real time. For each analyst question, we measured two features. The first is perplexity: how difficult the question is to predict from the management team’s prepared remarks. The second is topicality: how closely the question is connected to those prior remarks. We refer to questions that are high on both dimensions as curveballs.

In our study, questions with higher curveball scores were associated with several signs that executives feel unsettled upon hearing them. Responses were longer and more likely to be non-answers. Curveball questions were more likely to induce the CEO to step in and respond rather than deferring to the management team member with the relevant expertise. At the call level, curveball questions were also associated with larger absolute stock-price movements, larger absolute abnormal returns, and a greater likelihood that analysts changed their average recommendation. The pattern suggests that curveballs unearth novel information, prompting analysts and the market to reevaluate their understanding of the firm’s prospects.

We also examined the conditions under which curveballs were more likely to arise. Star analysts and analysts who were more central in professional networks were more apt to pose them than were their peers—suggesting they are likely to be found in the toolkit of the most savvy and experienced investigators. Curveballs were also more likely to be asked when analysts disagreed about a firm’s prospects, when a firm was difficult to classify categorically, and when an analyst was relatively new to covering the firm.

Asking Better Curveball Questions

Before we explain what managers can take away from our research, it is worth highlighting the research’s limitations. We acknowledge that the setting we study is unusual in that the evaluators (financial analysts) are especially knowledgeable and sophisticated actors, while managerial responses face tremendous scrutiny and can have immediate financial consequences. However, we demonstrate that our results extend to another context, U.S. Federal Reserve press conferences, in which evaluators are perhaps somewhat less sophisticated, even if the stakes are also quite high.

So, although we do not have direct evidence from other evaluative settings, such as job interviews or board meetings, we expect our framework to apply more broadly. In particular, it should be relevant when respondents are motivated to anticipate and prepare for evaluators’ questions, evaluators are sufficiently informed to ask questions that are difficult to anticipate, and the interaction allows for impromptu follow-up questions.

With that in mind, we offer seven pieces of advice for how interrogators can work curveballs into interviews.

Work from their narrative.

A curveball question is, by definition, relational: it is unexpected relative to some prior discourse. In other words, the same question can be predictable in one conversation and unexpected in another. So it is important to start with the respondent’s own account. What claims are being made? What assumptions connect them? In a strategy review, for example, management may be prepared to discuss its demand forecast but less prepared to explain how that forecast depends on a particular distribution partner. The opportunity lies in probing the unexamined link, not in choosing a more aggressive tone.

Ask for connections, not just facts.

Many follow-up questions are designed to elicit more specific numbers and dates or more concrete examples of claims that are being made. Such questions may be necessary, but they are also relatively easy to anticipate. A more revealing question may instead ask the respondent to connect the dots between disparate claims. How does the new pricing model affect the retention goal mentioned earlier? Where might standardization and local autonomy come into conflict? What does the hiring plan imply for the cost target? Such questions hew to the respondent’s account while requiring them to make logical connections that may not have been worked out in advance.

Change one consequential assumption.

Another source of curveball questions is to scrutinize a core assumption underpinning an argument and stress test whether the argument would still hold if the assumption were overturned. What would you do if demand grew half as quickly? Which part of the recommendation survives if the regulatory timeline slips? What evidence, if it were true, would cause you to reverse your decision? These counterfactual scenarios must, of course, be plausible and consequential. A far-fetched hypothetical question may be surprising, but if it is easy to dismiss as irrelevant, it is not a curveball.

Combine expertise with an outside view.

Our findings point to a useful tension. Curveballs were more likely to come from analysts with deep expertise but also from analysts who were relatively new to covering the firm. Managers can try to preserve both advantages. In hiring, this might mean including someone on the interview committee who understands the job candidate’s role but was not involved in formulating the interview guide. In a project review, it might mean asking a knowledgeable colleague from an adjacent team to examine the plan.

Use surprise judiciously, not to “perform” toughness.

Curveball questions need not be adversarial; rather, they can often aid the learning of both parties. In a leadership meeting, they might help uncover a hidden dependency that would otherwise pose a risk to successful implementation. In a coaching conversation, they can help an employee reconsider an assumption that is causing them to talk past another colleague. In many cases, it can help to make the purpose of a curveball question explicit. For example, one might say, “Because we know how important this decision is, I want to ask how your recommendation would hold up if the competitive landscape shifted significantly between now and implementation time.” Framed correctly, a question can be demanding, while leaving the respondent room to think—and without placing them under undue stress or pressure. In contrast, a question designed mainly to embarrass someone is unlikely to improve learning.

Time your questions.

For a question to be a curveball, it must be posed at the right time—specifically, when it is very much on topic with what an evaluator has just stated and is therefore difficult to dismiss or deflect. In contrast, evaluation targets who are media savvy or especially skilled at engaging with audiences often know how to parry an elegant, difficult-to-anticipate question that is off topic.

Change up the recipe.

Once a question becomes standard, respondents learn to prepare for it. The implication is not to build a permanent list of “best curveballs.” In fact, even following the steps above too literally increases the likelihood that a respondent will anticipate and begin to prepare for curveball questions an evaluator might try to pose. Effective evaluators therefore must constantly change up their recipe for surfacing difficult-to-anticipate questions.

The Age of the Curveball

AI makes the limitations of traditional “tough” questioning more consequential because it lowers the cost of preparation. Respondents can use generative tools to anticipate likely questions and rehearse answers. Generic “hardball” questions may therefore become easier to prepare for. So as preparation becomes easier, questions whose difficulty depends on the specific context may become more valuable.

The most revealing question is not necessarily the most aggressive or unusual. It is topical such that the respondent must engage, yet it is difficult to anticipate such that a prepared answer will not suffice. By probing assumptions, making connections, introducing plausible alternatives, listening adaptively, and continually refreshing their approach, managers can use curveballs to learn more, without making conversations contentious.

Copyright 2026 Harvard Business School Publishing Corporation. Distributed by The New York Times Syndicate.

Explore AAPL Membership benefits.