As a former marketing director and current professor, I have seen frequent exhortations in writings and speeches for business leaders to be more data-driven when making decisions. While I am very much in favor of decisions being grounded in data, data-driven decision-making may be a bad idea for at least three reasons.

Risk of drowning in data

Data-driven decision-making may be a recipe for organizational paralysis if taken too far. There is a lot of data in the world, and more so every second. Statistica forecasts that raw data will almost triple from 2020-2025 (Statistica 2021). Subjecting all decision-making processes to data analysis will eventually result in the inability to make any decision, no matter how fast our computers and smart our AIs.

Ignoring other sources of knowledge

Decisions are never made directly upon the data. They are based on beliefs about the future and how the world works. Beliefs used in decision-making may be grounded in data, but they also are based upon other things – experience, theories, and instincts. This is not always a bad thing, as some might imply.

For data to change current beliefs, it first needs to be analyzed to sift the signals from the noise. Then, new findings also have to be reconciled with the previous beliefs. Better decisions will not result if one buries current beliefs in avalanches of big data or overrides them with algorithms. On the contrary, better decisions require smarter integration of beliefs of different quality and provenance with the results of fresh analyses.

Overinvesting in information technology

The phrase data-driven decision-making may cause businesses to misallocate their investments in improving decision-making quality and effectiveness. The current emphasis to improve data collection and advanced analytics leads organizations to over-invest in big data projects and high-tech analytic platforms, and under-invest in the human talent still required to make decisions based upon the results. This leads to disappointing returns, and outright failures, of many data analytic projects.

How to manage a belief-based decision-making process

In my corporate career, there were many times when the research department would present a new study as if it was a completely new revelation. It seemed at times that we were being told to ignore wisdoms accumulated over 30 years of experience in the category.

On the flip side, there were times when I was the one trying and failing to challenge prevailing management wisdoms with new research. The resistance to accepting the new findings was particularly strong when someone’s personal status in the organization appeared threatened by the results, or they feared that their judgement was being replaced by an algorithm. As a result, new analytic findings were often ignored and bad decisions were made.

A greater sensitivity to the role of beliefs would have caused me to do at least two things differently. One, before launching analytic projects, I would have spent more time working with the analysts and management to identify and quantify their existing beliefs. Doing so upfront would have helped to clarify the quality and quantity of information required to change minds.

Two, I would have pushed the use of Bayesian statistical methods rather than the Frequentist methods that were commonly used. Bayesian methods explicitly incorporate prior beliefs into the data rather than largely ignore them, as Frequentist methods tend to do. This would have enabled the decision-makers to revise their beliefs incrementally rather than be asked to make all or nothing choices.

Not only is a belief-based model of decision-making a better fit with how decisions are actually made, it can incorporate a statistically sound approach for integrating gut-level instincts with advanced analytics. Belief-based decision-making lacks the alliteration of data-driven decision-making, but it offers a basis for decision-making that is informed by data but not a slave to it.