Advertisement

Whether it’s a job seeker evaluating offers, an individual buying or selling a home, or a business assessing bids from buyers or suppliers, people routinely search for the best option without knowing in advance the range and odds of the options they'll encounter.

This fundamental challenge is the starting point for new research by David Brown, the Snow Family Business professor at Duke University’s Fuqua School of Business and Fuqua PhD candidate Cagin Uru.

Related stories

In their paper, Data-Driven Sequential Search, Brown and Uru revisit a classic economic problem known as Pandora’s Problem—where decision-makers choose between selecting the best option seen so far or continuing to search at a cost in hopes of finding a better option.

In this setting, it’s assumed that the decision-maker knows how options are distributed—for instance, that auction bids or product prices follow a certain pattern.

But what if you don’t?

This challenge arises frequently in dynamic markets, where conditions can change rapidly and meaningful data may be difficult to obtain, Brown said.

“People often search without having any idea what the next alternative might look like,” he said. “In some real-world settings, it’s difficult to know how valuable future opportunities might be. Essentially, you are searching in the dark.”

Rather than relying on assumptions or predictive models, Brown and Uru focused on strategies that operate in the absence of prior beliefs, and commit in advance to search for only a fixed number of alternatives—and pick the best one.

The researchers found that their approach, despite its simplicity, guarantees strong performance—often close to the performance of an oracle decision-maker who has full knowledge about the range and odds of alternative values.

Why Simplicity Wins

The key to the researchers’ strategy lies in choosing the right amount of search effort: rather than developing complex adaptive algorithms, simply commit to searching a fixed number of options, and then pick the best one.

“Don’t try to adapt on the fly. Just commit,” Uru said.

For example, suppose you’re trying to sell a product, such as a rare vintage of wine. Rather than using received offers to continuously infer the highest price the market may offer, you can simply commit to receiving, say, five offers, then take the best one, he said, while pointing out that this approach doesn’t require any detailed information about wine ratings, historical trends, or market conditions.

Brown and Uru show that in many real-world price and value scenarios—when the underlying distribution of values has certain structural properties—their commitment-based strategy performs favorably compared to the oracle.

“No search strategy relying solely on data, no matter how complex, can do much better than our strategy,” Uru said.

In fact, in some settings, such as when search costs are low or alternative values are equally likely, the researchers show that this simple approach is the optimal one.

Testing on Price Data from Art and Wine Markets

Brown and Uru tested their theory on two specialty goods with different variability in prices—vintage wines and art pieces—to demonstrate how their strategy works across a range of market conditions.

In both cases, their fixed commitment strategy came close to the performance of the oracle with full information about the odds of different markets. Notably, this was true even in the art piece markets, where prices are highly unpredictable.

“The key takeaway: adaptive decision-making may not add much or any value when you don’t know what’s coming next,” Uru said. “Sometimes, structure—even rigid structure—can outperform more dynamic strategies that adapt to noise.”

Implications for Businesses in the Age of AI

This research has implications far beyond art and wine, Brown noted. Any business facing high uncertainty and limited data—from hiring decisions to venture capital investing to consumer pricing—could benefit from a simple, disciplined approach to sequential decisions, he said.

Business leaders should consider setting clear exploration budgets, he said, and commit to reviewing a fixed number of options before making a choice—especially when the value of future options is unknowable.

For consumers and professionals, Brown recommended recognizing that chasing “just one more” option can backfire in uncertain environments, while a good-enough strategy may beat more complex, potentially dynamic, search strategies.

More broadly, the researchers believe their work enhances our understanding of when advanced analytic techniques, including those based on AI, can improve the quality of data-driven decision-making.

“For some decisions, you may not need to rely on sophisticated algorithms to get good results,“ Brown said. “Simple decision rules can match or exceed complex methods under uncertainty, while at the same time being easier to explain and interpret, and requiring essentially zero infrastructure.”

This article has been reproduced with permission from Duke University's Fuqua School of Business. This piece originally appeared on Duke Fuqua Insights

First Published: Oct 28, 2025, 18:27

Subscribe Now
Advertisement
Advertisement
Advertisement

Latest News