“The first principle is that you must not fool yourself—and you are the easiest person to fool.”
— Richard Feynman

I recently listened to a podcast about Richard Feynman, the Nobel Prize-winning physicist who helped develop quantum electrodynamics and later served on the commission investigating the Challenger disaster. What made Feynman interesting wasn’t just that he was brilliant. It was that he was relentlessly skeptical, especially of his own conclusions. He believed that questioning assumptions was more important than defending them.

That quote has been rattling around in my head lately.

I’ve spent more than twenty years in research. During that time, we’ve moved from phone interviews and desktop surveys to mobile devices, online panels, AI summaries, social listening tools, transaction data, and enough dashboards to wallpaper an office. We have more information available today than at any point in history.

Oddly enough, I find myself less certain than I used to be.

The challenge isn’t necessarily the analysis. It starts with the inputs.

In B2B research, we often survey restaurant operators, distributors, chefs, purchasers, and executives. We assume respondents are who they say they are. We assume they’re paying attention. We assume they’re thoughtfully considering the questions. Most are. Some aren’t. The problem is that we rarely know which is which.

An operator might complete a survey while unloading a truck, waiting for a meeting to start, or scrolling through emails between lunch and dinner rushes. They may answer honestly, but quickly. They may estimate. They may tell us what they aspire to do rather than what they actually do. That’s not deception. That’s human nature.

Then comes the part that has bothered me more as the years have passed.

What do we do with answers that don’t make sense?

Research professionals are trained to identify outliers, remove suspicious responses, and clean datasets. Sometimes that’s exactly the right thing to do. But sometimes I wonder whether we’re throwing away noise or throwing away reality.

If ten operators tell us something unexpected, are they bad respondents? Or are they seeing something we aren’t?

The answer often depends on who’s reviewing the data. One researcher may see an invalid response. Another may see an emerging trend. A client may see a conclusion they dislike. Human judgment enters the process long before the final chart is built.

AI doesn’t solve this problem.

AI can summarize information faster than any researcher I’ve ever met. It can identify patterns and inconsistencies. What it cannot do is determine whether the underlying information reflects reality. If the inputs are flawed, AI simply processes flawed information more efficiently.

The longer I work in this industry, the less I believe research is about finding answers and the more I believe it’s about asking better questions. The data matters. The methodology matters. But curiosity matters too.

Feynman understood that the greatest danger in science was fooling yourself. I sometimes think the greatest challenge in modern research is the same one. Not because we have too little data. Because we have so much of it.

 

Tim Powell is a Principal with Foodservice IP, a Food-Away-From-Home consultancy. To learn more about FSIP’s Management Consulting Practice, click here.

Like the content? Sign up to receive our communications.