Physicist Niels Bohr once mused that an expert is someone who has made all the mistakes that one can make in a narrow field.

That's not a surprising sentiment coming from a scientist. Science is based on exposing ideas to the rigor of real-world testing. It's an antidote for wishful thinking, superstition, unjustified enthusiasm and overconfidence. Failure, in this way of thinking, is constructive because it brings you closer to the truth.

That said, even in tech circles where you'd think there would be reverence for the scientific method, using controlled randomized testing is severely underutilized. There are exceptions though. Airbnb claims to run some 700 experiments a week, and CapitalOne is known as a pioneer in business experimentation, but more often, companies do way too little real-world testing. One of the main reasons is a very human one - they have been trained to think that they are supposed to know what will happen.

What is business experimentation?

How do I know that few companies are disciplined about experimentation? Because businesses make mistakes all the time. Even companies that we think are infallible, like Netflix, make huge mistakes. Recall that back in 2011, Netflix famously angered its customers by splitting its DVD-by-mail and streaming businesses and raising prices for both.

In that case and others, Netflix would have benefited from testing this idea with a small, but representative portion of the market. (They later perfected their pricing as my colleague, Kyle Poyar, writes about here.) Then, at least it would have known how the notion would go over with consumers. There are other recent misfires as well: Juicero, maker of the $400 Internet-connected juicer maker that raised $120 million in venture funding before shutting down last year; Google Plus; the Amazon Fire phone. The list goes on. That's not to say that failure is bad - far from it. But such failures might have been realized on a smaller scale until the idea was either refined or discarded. That's not always the case, but it is often enough.

Readers at this point might retort that they do a fair amount of A/B testing for marketing messages. My reply is - do you, really? Based on the rhetoric, it might seem like 100 percent of companies are executing A/B testing for marketing, but in reality it's more like 2 percent.

As a result, marketing budgets are off. A typical marketing budget for this year is going to pre-assume that if you put in a certain amount of resources for marketing that you're going to be able to get a certain amount of results. But, particularly with young companies, that's not accurate because marketers haven't experimented with enough marketing channels or marketing messages or other variables to be able to understand what the relationship is between what they do and the behavioral response from their audience.

If companies aren't doing enough web-based experimentation, that's a bad sign because that's the easiest type of experiment to run. It's easy to expose a representative sample of 50,000 or so prospects on the web to a marketing message

In the field, it's also rare to find companies using the scientific method and comparing their randomized selection of consumers to a control group and controlling for variables and so forth. If they try to experiment, companies often do it half-heartedly and the results are just statistical noise that can lead them to make bad, or at least random, decisions.

Why experimentation is hard

Experimentation is the cornerstone of innovation. No matter what the operation is, no matter what the functional area is and no matter what the operation is like, there are always opportunities for it to get better. The only way to get better at the fastest way possible is to run experiments.  The more the better. The more good ideas that are tested, the better.

Doing so isn't easy. If you run your organization really lean then you barely have time to do your transactional work, much less run experiments on top of that. But a bigger obstacle is the fear of failure. Our educational system has drilled into us that there is a right answer to every problem and if you do whatever is right, you'll get the right answer. Even in business school, students are expected to be able to predict the answer. When you don't - when you're wildly incorrect - it's humbling.  

The key is to change your mindset and think about the outcome of a successful experiment being a quick, inexpensive, statistically valid result rather than the outcome being any specific result.  Then, figure out how to scale up the number of experiments that you run at any given time. Finally, put the better ideas into production.

So change your mindset and look at failure the way scientists do - as necessary, helpful and the only true path to mastery and expertise.

Published on: Mar 22, 2018