Companies that don’t automate these decisions will lose money and quality.

Some decisions can literally change people’s lives. A physician’s interpretation of a scan, a judge granting or denying joint custody, or a bank manager approving a loan that enables a new venture can determine how our lives—and those of the people around us—unfold.

In all three cases, outcomes depend far more on luck than we’d like to admit. Luck in who the physician, judge, or analyst happens to be—and luck in whether that person had a bad night, is dealing with a personal crisis, or is simply having an off day. This is what we call noise: the unwanted variability in professional judgments that should be identical.

In the corporate world, this kind of error is costly and can generate significant losses. Daniel Kahneman—psychologist and Nobel laureate—illustrates this in his book co-authored with Olivier Sibony and Cass Sunstein. An analysis at an insurance company found 55% variability in premium quotes for the exact same case. In other words, under identical conditions, underwriters’ decisions varied by 55%, harming the company and some customers.

This variability appears in performance evaluations, investment decisions, M&A, and budgeting. It is an expensive problem at both individual and organizational levels—and only recently has it reached many executives’ radar.

Errors in Judgment

There are two distinct families of decision errors: bias and noise. Bias (discussed in prior columns) refers to predictable, systematic errors that generally push in one direction. Noise is a statistical measure of the unwanted variability in outcomes. Consider the scales you find in pharmacies or clinics: if, on average, a scale reads consistently too high or too low, it is biased. If the same person steps on the scale repeatedly and gets disconnected readings—three kilos more, two less, then 1.5 more—that is noise. Bias is the average error; noise is the dispersion of errors.

Noise is not only costly—it is often invisible. Few people in companies actively think about it. Human judgments vary widely and depend on factors like context and mood. Try this: on vacation, describe the taste of a wine and rate it, then repeat the exercise in nine months without checking your notes. Odds are your description will differ. The good news is that variability can be measured and reduced. Many corporate decisions improve simply by examining how they’re made, structuring them more rigorously, digitizing parts of the process, using algorithms, or deploying existing technologies.

Some companies pay close attention to decision processes—and, as expected, achieve better results. China Mobile, Pfizer, and Chevron have won the Raiffa-Howard Award (named after Howard Raiffa and Ronald Howard of Harvard and Stanford), granted annually by the Society of Decision Professionals to recognize consistent, organization-wide excellence in decision making. These firms use processes and tools to decide better—and they also build a culture and capabilities that empower people to make decisions the right way.

It’s important to distinguish recurring decisions from one-off decisions. Recurring decisions—where consistent answers are expected—are prime candidates for programmed algorithms that reduce noise by replacing human judgment. Algorithms don’t have bad days, aren’t swayed by headlines, and don’t depend on external factors. Under identical conditions, they deliver the same response every time—and with AI, they learn faster than humans. Companies that don’t automate these decisions will lose money and quality.

“Next-best-action” engines, like those used by Netflix to recommend the next series, are a strong example that many brick-and-mortar firms are now emulating. Similarly, when large data volumes are available, experimentation becomes a decisive advantage. Booking has cultivated a test-and-learn culture where anyone can run the experiments they need without seeking permission—resulting in roughly 25,000 online tests per year.

For infrequent decisions, the playbook is similar. Exploration and production companies like Repsol and Exxon have long archived decision analyses for later review to inform future, comparable choices. This post-mortem analysis examines how decisions were made, the methods and tools used, and helps improve judgment as organizations learn.

Turning down the volume on this invisible noise isn’t easy. Continuing to operate at today’s decibel level is far riskier—both for your next medical diagnosis and for your business results next quarter.

This is a translation of the article originally published in Diario Expansión.
Read the original piece: https://bit.ly/3CgWzf3