The following is based on several weeks of conversation with JP Castlin which explored issues of experimentation — discussion which led to a first and second joint article in the Strategy in Praxis newsletter and one I authored about emergent goals for experimentation. (Some Strategy in Praxis content is reserved for subscribers.)

The following is derived from these discussions though the opinions expressed here are my own. So while JP deserves credit for contributing to what is useful and good, he deserves no blame if my writing has gone astray.

Rory Sutherland tells a story about bees. Apparently, orders are issued within a hive by a bee sergeant doing what is called a waggle dance. Casual outside observers clearly see worker bees as they head off following orders to harvest pollen.

This isn’t the whole story, though, What casual observers can’t see is what scientists have found — that a portion of the bees do not follow the orders of the hive — they rebel. What portion? That’s not entirely clear but I generally imagine 20% or 30%.

And what do they these rebellious bees do? This story is not some apian version of Ferris Bueller’s Day Off as, last I checked, bees can’t drive Ferrari’s. Instead, they explore around the neighborhood of the hive — exploration which discovers new sources of pollen as well as locations where the hive could move in an emergency.

This is all fine and good. But the exceptionally critical point is that BOTH groups of bees are necessary for the health of the hive. Without rebellious bees the hive would be less resilient and less likely to survive.

Exploitation v Exploration

All complex adaptive systems are also only healthy when exploitation of what is known (e.g. pollen sources) is balanced with time exploring things critical to future health. It is for this reason that JP Castlin talks, then, about the need for continual “safe-to-fail” experimentation in business as all companies are complex adaptive systems. All companies, then, require continual exploration if they are to be healthy. Of course, he calls these “safe-to-fail” experiments because they must be affordable enough — or constructed in safe enough ways — that they don’t sap the company’s ability to survive.

Unfortunately, the mere mention of experimentation leads to one of two responses.

• Many might imagine aggressive exploration as expressed in the silly Mark Zuckerberg admonition to “move fast and break things.” We are tremendously fortunate that NASA viewed their mission as one where “failure is not an option.” Neither approach is right in all situations. Zuckerberg’s world of software changes was fundamentally quite safe so an extreme quote might work. NASA’s missions sending humans into the vacuum of space and other great dangers required failure not be allowed to be an option.
• Other companies find the idea of continual experimentation to be incredibly uncomfortable because it opens the door to all four types of uncertainty JP and I discussed in the second article. Making matters worse, too many companies respond to these uncertainties by adding bureaucracies hoping to control risk — bureaucracies which ensure exploration fails.

Consider, again, the bee hive. When the rebellious individual bee does not follow bureaucratic hive rules they don’t follow an alternative set issued from an R&D bureaucracy — their efforts are not overseen by risk managers. Instead, it appears the rebellious bee is self-directed to explore what it perceives important (according to some bee-level version of “perception”).

A first great error companies make is to attempt to rid themselves of employees who sometimes rebel. This error, then, compounds when they decide to run experimentation with a bureaucracy.

To be clear, rebellious bees don’t head out to do “just anything” as if apian Ferris Beullers. They have their own deep understanding of the hive and the world around the hive. It seems reasonable to assume they make wise use of their time based on clearly understanding the business of the hive. So, too, in companies. Managerial hubris may want to believe employees are mindless peons (replaceable by AI) but they are not. Even the youngest and least experienced employees bring a natural sense of what is important — a sense which develops significantly as they mature within a business.

Instead of directing exploration, then, it should be overseen with a type of lassez-faire management enabling individuals and their teams to explore what they perceive as critical without having to continually justify their choices. Thus, companies should rely on those employees (perhaps 20% to 30%) who rebel against rigid management yet comprehend what matters to the company business. This is not to open a free-for-all. Rather, companies should minimize the management interference which kills the value of experiments while ensuring employees work wisely and report what they are doing.

Companies far prefer, though, that employees be passive, rule following bees unable to see what William Whyte (The Organization Man, 1956) observed:

[T]his the administrator cannot conceive…that a man [or woman] can dislike the company…and still have made a net contribution to the company cash register infinitely greater than all of his [or her] better-adjusted colleagues put together.**

Exploration does not succeed if companies dictate its goals.

Why Explore and Experiment?

What types of very good things happen as we explore? Notice I did not suggest we define “goals.” The term “goal” is dangerous with experiments as the idea of a “goal” today means aims which are “specific and precise.” The Oxford definition, though, tells us a goal is merely “the object of a person’s ambition or effort; an aim or desired result”.

Within Oxford’s looseness, then, many experiments hope to find out whether it’s true that “if we do X then Y will result.” Uncertainty starts with whether we can do X (not all things can be done). Uncertainty continues, though, with an unspoken truth — there are always many ways to “do” X. So the most critical question is whether there is way to do X where “Y” somehow results. Despite this, companies experiment believing they need to determine whether “X” leads to “Y.” The most critical result from an experiment isn’t “whether” but “in what ways can we act such that X leads to Y.” THAT is serious learning.

A sometimes more productive approach to experiments tries new actions to see what happens answering “what will happen if I do X?” These experiments can be highly productive because they learn how (a) different ways of doing X lead to (b) different end results.

Experiments also reveal limits around our work and help us discover opportunities made possible when what we believed were limits turn out not to be limits. Note that best practices are often limiting, rigid rules which become false limits. Experiments can show us where those rules are wrong.

As we explore, then, three broad values might emerge which are critical for our companies. Yet we cannot, in advance, determine which our experiment will discover nor, ultimately, how a specific experiment may be valuable. Thus, companies need to experiment to learn what can be learned — and from that learning begin to discover ways to proceed with great strength.

Resilience from Continual Experimentation.

Experimentation must not be isolated only to R&D efforts. Companies need to pursue a continual thread of experimentation in every department, project or operation. As an example consider a company’s inbound call center. Such a call center should continually experiment with the ways in which they answer, and process, calls. These experiments may discover new approaches. Far more likely, they will discover approaches which fail in many (or most) situations but are sometimes quite useful.

Continual experimentation of this type helps departments be resilient. More critically, it builds company-wide ability to adapt. When experimentation is isolated only in R&D then the company will never reap the benefits of what that R&D might find. Why? If an innovation is important, all other departments must adapt their work.

Many brilliant inventions — perhaps like some notable ones from Xerox PARC — succeed only outside the company because the rest of the company is incapable of adapting to make them work. I once wrote a series of blog posts with four primary and seven secondary diseases known to kill innovation. These posts noted how a dictatorship of KPIs, best practices, and formal risk management often kill innovations. Why? If something is truly innovative, KPIs, best practices, and risk management must change or the innovation fails. These rigid controls are put in place, after all, to ensure stability and not to enable adaptation.

Consider, instead, how early- and mid-20th century AT&T paired the R&D from Bell Labs’s with Western Electric who turned discoveries into products. AT&T, then, involved managers at Regional Operating companies (like my father) who regularly toured both Bell Labs and Western Electric so their plans leveraged what was coming.

Product/Service or Process Related Innovation to Avoid Revenue Loss.

A valuable result from experimentation is not always an important innovation. Often experimentation is needed to discover ways to maintain current business as markets, industries, and customers change. In other words, these results avoid revenue losses. I confess to being a bit torn with this type of experimentation. Often what is discovered requires great expense with only relatively low reward yet companies must make the investment. That said, at other times companies embrace these experiments too quickly because they seem certain to not fail when compared with innovations leading to future demand.

A simple example of avoiding revenue loss might have a cruise ship replace its deck chairs. This might be expensive (have you checked the costs of such industrial grade furniture lately?) but also won’t increase the number of passengers — it only keeps from losing passengers.

Another example is found at Kodak with their 1980s introduction of 126 Instamatic Cartridge film. Already dominant in the film market, little upside was possible from this product (though there may have been benefits of which we are not aware). Yet the introduction must have been quite expensive.

Product/Service or Process Related Innovation to Build Future Demand.

The ultimate gold-standard result of experimentation is an innovation which can dramatically affect future demand for company products and/or services. Companies must always explore if they are to have sufficient demand for products and services in the future. In one draft I called these “big experiments” because companies cannot survive without their discoveries. But, to be clear, this exploration is not always costly in time, energy, or money. Sometimes, inexpensive but high risk exploration uncovers potent new sources of future revenue.

Experiments of this type, though, are always less certain in their results than loss avoidance experiments. Hence, savvy companies need to stage such experiments to ensure spending increases only as returns from experimentation sufficiently justify the next cost (though “sufficient” is a vague term only possible to judge — not to calculate). I have found two important truths around uncertainty in these experiments.

• In their earliest phases, paths to superb new opportunities are often murky and ill-defined. Critically, they rarely shout out their potential. Yet if we pursue only those opportunities clear in the beginning we will not find the demand needed in the future.
• Uncertainty will never be fully resolved as we develop future demand. Instead, it remains present until rollout proves the potential. All along the path, then, work must focus on learning without trying to solve the whole problem all at once.

As a side note, companies should rely more often on thought experiments as early steps. Unfortunately, many businesses have such strong preferences for thoughtless action that they eliminate the individuals capable of effective thought experiments. Such individuals have unique abilities to imagine the inventive exploration surrounding a path while also judging whether it should be pursued. It’s a tricky job which I particularly enjoy.

What About Performative Experiments?

Making our lives in business far trickier, experimentation has been co-opted to serve the process of seeking investment and satisfying shareholders. I will call these “performative experiments” as their goal is to prove to possible money sources that the company is experimenting. Of course, such “experiments” don’t need to be productive learning — they must only drive investment or shareholder satisfaction. In part this is a result of incredible hype about experimentation.

As an example, I have come to believe many “radical” ideas announced by Amazon primarily offered shareholders an excuse for unprofitability. It worked. During this time I was regularly told by shareholders that Amazon WAS superbly profitable but spent all the profits on experiments. I was quite skeptical at the time and continue to be. It’s doubtful that Amazon ever had the claimed profits. Instead, they funded these experiments through their vaunted cash flow — cash essentially borrowed from vendors through extended payment terms. Their performative approach, though, convinced investors to give Amazon freedom to ignore profit as long as share prices increased. It also required that sales continually increased for cash flow to also continually increase.

Companies claiming performative experiments need to be wary. These experiments often do not result in useful learning because that is not their focus. These experiments also usually fail. For example, there had been useful and savvy exploration of connected eye-wear by superb engineers at Google. Yet its executive founders took over and turned the project into a performative announcement which created the Google Glass failure. Today I’m guessing Walmart’s recently announced “Grab and Go” shopping in SamsClub is also performative — there is little evidence it can have any meaningful impact on their business. And Elon Musk’s long list of claimed invention is also a list of performative experimentation as seen in this Wired Magazine article.

The Necessary Company Rebels

It’s important to qualify the idea of rebellion. Political rebellions, after all, involve organized groups using violent means attempting to create change by burning buildings and killing kings or politicians. They also lead to authorities responding in kind with equally violent actions.

Our concern is far different as rebels within companies are often quite decent and unlikely to know anything about Molotov cocktails. Instead, companies perceive any individual ignoring a corporate rule (or rules) to be a rebel. Companies need more such rebels. When visiting companies I consider it a bad sign if everyone loudly states absolute loyalty to rules and best practices — a clear “tell” of a company with disaster not too far away in its future.

Healthy companies, instead, always have some degree of chaos — even potential mutiny. I particularly love this quote from automotive executive Bob Lutz about his time at Renault:

…we often had chaotic meetings, where some of my direct reports engaged in behavior that an outsider would have termed insubordinate bordering on mutinous. But that’s when we had the clearest communication and surfaced ideas and opinions devoid of the buffering wads of tissue paper designed to avoid hurting someone’s feeling.***

No company, it turns out, can be healthy unless some portion of its work force focuses on what is important while ignoring unrelated and interfering corporate orders. Bureaucracies, though, hate how this idea rejects their idealized belief that orderliness is next to godliness.

Until next time, be well and go forth to rebel in savvy ways. And May the Force Be With You.

©2025 Doug Garnett — All Rights Reserved

Through my company, Protonik LLC, I consult with companies as they design and bring to market new and innovative products. I am writing a book exploring the value of complexity science for driving business success. Protonik also produces marketing materials including documentaries, websites, and blogs. As an adjunct instructor at Portland State University I teach marketing, consumer behavior, and advertising.

You can read more about these services and my unusual background (math, aerospace, supercomputers, consumer goods & national TV ads) at www.Protonik.net. Roughly once a month, Shahin Khan and I discuss current issues in marketing on our podcast The Marketing Podcast available on Google, Spotify, the OrionX website, and Apple Podcast. 

** The Organization Man, William Whyte, Simon and Schuster, ©1956, p141

***Car Guys vs Bean Counters, Bob Lutz, ©2011, Apple Books edition, p51