INNOVATION&

INNOVATION&

The Anomaly Is Not the Enemy

Why operational excellence and innovation require different responses to deviation

Yetvart Artinyan's avatar
Yetvart Artinyan
Nov 17, 2026
∙ Paid

I recently heard someone say in a business conversation, “This is an anomaly. We need to eliminate it—or at least reduce it.”

That sentence reminded me of Clayton Christensen and the handmade sign outside his office at Harvard Business School:

Anomalies Wanted.

Christensen was not celebrating failure or disorder. He wanted anomalies because they exposed situations in which accepted explanations could no longer account for reality. An unexpected observation might be a measurement error or an irrelevant exception. But it might also reveal an incomplete theory, a misunderstood causal relationship, or a circumstance nobody had considered.[1]

This distinction matters beyond academic research. In established operations, deviations often indicate something that needs correcting. In innovation, however, the standard itself may still be wrong. Eliminating deviations too early can eliminate the evidence needed to discover something better.

The anomaly is not necessarily the problem. Sometimes, it is the first indication that our definition of normal is incomplete.

The free section of INNOVATION& is offered freely, and you are welcome to keep reading it that way. If these articles help you make better decisions in your working life, please consider becoming a paid subscriber. Researching and writing them takes time, and paid subscriptions provide the income I need to continue. Your support also helps keep part of this work available to everyone.

When eliminating deviations makes sense

In a sufficiently stable environment, an organization understands its processes, expected outcomes, and principal risks. It can establish standards, measure deviations, identify causes, and improve performance. The underlying logic is straightforward:

Standard → Action → Deviation → Correction → Compliance

This approach is appropriate for production processes, financial controls, safety procedures, and repeatable services. When a validated process produces an unacceptable result, the organization needs to investigate and correct it. That investigation may reveal that the process or standard itself needs revision, but maintaining reliable performance remains the immediate objective.

Over time, successful correction reinforces the standard. Processes become more predictable, behavior becomes routine, and routines establish what the organization considers normal. This is operational excellence doing its job.

The difficulty begins when organizations apply the same logic in environments where they do not yet know what normal should be.

Follow me on LinkedIn

Innovation begins before a reliable standard exists

Innovation involves assumptions that have not yet been sufficiently tested. An organization may not understand the job customers are trying to accomplish, which solution they will adopt, what they are willing to pay, or whether the proposed business model can become economically viable.

Under these conditions, an unexpected result cannot automatically be classified as an error. It might be random variation, an overlooked variable, evidence that an assumption is wrong, or an opportunity nobody anticipated. Determining which interpretation is justified requires investigation.

The learning logic is therefore different:

Assumption → Action → Anomaly → Inquiry → Revised Explanation → Next Test

This resembles James March’s distinction between exploitation and exploration.[2] Exploitation improves what an organization already knows how to do, while exploration searches for alternatives whose value and consequences remain uncertain. Both are necessary, but they require different responses to unexpected outcomes.

When every deviation is treated as an error, the organization risks defending its existing explanation of reality rather than investigating whether that explanation remains valid.

That does not mean every anomaly deserves attention or that preventable harm should be tolerated. A dangerous event may require immediate containment, a regulatory violation must be addressed, and a harmed customer does not become an interesting experiment. But containment and interpretation are different decisions. We can stop the consequences of an event without prematurely deciding what caused it or what it tells us.

Organizational learning requires questioning not only actions but also the assumptions and governing rules behind them.[3] Sensemaking begins when expectations are interrupted and people attempt to explain what they have observed.[4] An anomaly creates precisely such an interruption.


Facing a decision under uncertainty?

Before you commit more capital, credibility, or resources, let’s examine what you know, what you assume, and what evidence would justify your next move.

I work with founders, executives, and teams on decisions about what to test, scale, or stop.

Available online worldwide and in person across Europe.

Request a 20-minute introductory conversation.

Reach out


Innovations sometimes begin with unexpected observations

History offers numerous examples of discoveries that emerged from observations inconsistent with expectations. Penicillin, X-rays, the microwave oven, PTFE, Post-it Notes, cyanoacrylate adhesives, the implantable pacemaker, Scotchgard, vulcanized rubber, and sildenafil all feature unexpected observations in their development histories.[5–14]

These stories are often simplified into accidental breakthroughs, as though an unexpected event somehow produced an innovation by itself. Yet the observation was only the beginning. Someone still had to recognize its significance, investigate the underlying mechanism, establish the conditions under which it could be reproduced, and connect it to a useful application.

The relevant sequence is not simply accident followed by innovation. It is unexpected observation, inquiry, reproducibility, application, and eventual adoption. Most anomalies will never become innovations, and being unusual says nothing about commercial value. What matters is an organization’s ability to recognize which observations deserve investigation before they disappear into routine correction.

Your organization is already an anomaly-sensing network

Companies encounter unexpected evidence every day, often through people whose primary responsibilities have little to do with innovation.

Salespeople hear why apparently ideal customers refuse to buy. Customer service teams discover unexpected uses of existing offerings, while help desks encounter workarounds that product teams have never considered. Marketing sees messages perform differently from predictions, finance notices unusual payment behavior, and procurement encounters supplier capabilities that do not fit established assumptions. HR observes behavior that challenges explanations of motivation and retention, while legal and compliance teams see emerging business models collide with rules designed for older ones.

These observations matter because they occur where organizational assumptions meet reality. Yet many companies organize these functions primarily for throughput, control, and escalation. A lost sale becomes another lost sale, a workaround becomes a support ticket, and an unusual request becomes an exception to be processed.

The organization records the event without necessarily preserving the assumption it might contradict. As a result, the problem may not be a lack of anomalies but the absence of a system for recognizing, connecting, and investigating them.

When I saw the process as the anomaly

Earlier in my career, I worked for a telecommunications company where preparing a customer quote required a technical feasibility study lasting two to four weeks. One evening, I noticed that several competitors already had infrastructure in a building we were evaluating. Why calculate an entirely new connection when existing infrastructure might be available?

That observation led me to develop a tool with a colleague, using competitors’ building databases and fixed-price offers. We reduced quotation times from weeks to minutes or hours, with aggregate pricing deviations of just 2 percent, later reduced to 1 percent.

Yet the approach initially met resistance from engineering and fellow account managers because it challenged established procedures. The tool was eventually adopted company-wide and remained in use long after I left.

The experience taught me something beyond the trade-off between quotation accuracy and response speed. What I considered an inefficient process, the organization considered normal. What I proposed as an improvement, the organization initially considered a deviation.

And sometimes the person questioning the standard becomes the anomaly. Challenging established practices can also challenge expertise, authority, and professional identity. Research on psychological safety and employee voice shows why people may withhold observations that contradict prevailing assumptions.[18–20]

An organization can claim to value innovation while discouraging the very people who question its definition of normal. When speaking up becomes riskier than remaining silent, the anomaly disappears from the conversation, not from reality.

Today’s innovation is often yesterday’s unwelcome anomaly and tomorrow’s unquestioned normal.

Why AI may make the problem worse

AI introduces an interesting complication. It can identify unusual patterns across operational data, customer interactions, and market behavior at a scale that would previously have been difficult to achieve. That capability can improve both operational reliability and the discovery of unexpected relationships.

However, detecting an anomaly and interpreting its significance are different tasks. Systems designed to identify departures from established patterns can be extremely useful in stable operations. Under uncertainty, those same patterns may reflect assumptions that have never been sufficiently tested.

An organization that uses AI primarily to classify unusual events as errors, fraud, noncompliance, or irrelevant noise may become highly effective at reinforcing its existing understanding of normal. The technology improves detection, but the decision logic following detection remains unchanged.

The important question is therefore not whether AI can recognize deviations. It is whether the organization uses those observations to protect an established model or to investigate where that model might be incomplete.

More data alone cannot settle that judgment.


Interested in working together?

Reach out


Stop celebrating failure. Start investigating signals.

The familiar advice to embrace failure is not particularly helpful. Some failures are preventable, repetitive, or unnecessarily expensive. Others reveal something important about assumptions that could not have been resolved without testing them.

Research on error-management cultures, learning through failure, and intelligent experimentation supports distinguishing between failures and their potential learning value.[15–17] The objective is not to produce more failure or to investigate every unexpected result. It is to respond intelligently when observations contradict expectations.

Operational excellence asks how deviations from validated standards can be reduced. Innovation asks what deviations might reveal about the validity of assumptions, explanations, and business models.

Confusing the two creates predictable problems. Treating every operational error as an invitation to explore undermines reliability, while treating every unexpected result in innovation as an error to eliminate undermines discovery.

The leadership task is not to choose between control and curiosity. It is to understand what is sufficiently known, what remains uncertain, and which response is justified by the consequences.

The next time an unexpected result appears, the first question should therefore be whether we are protecting a validated standard or protecting an assumption from evidence.

Sometimes the appropriate response will be to correct the deviation, and sometimes it will be to ignore it. But occasionally, it may be worth asking:

Hello. Who are you, and what are you trying to tell us?


The free section of INNOVATION& is offered freely, and you are welcome to keep reading it that way. If these articles help you make better decisions in your working life, please consider becoming a paid subscriber. Researching and writing them takes time, and paid subscriptions provide the income I need to continue. Your support also helps keep part of this work available to everyone.


Thanks for reading INNOVATION&! This post is public so feel free to share it.

Share


Start paid learning and application section


Turn an Anomaly into a Decision

Recognizing that anomalies may contain valuable information is relatively easy. The harder part is deciding which observations deserve further investigation, how much evidence is sufficient, and when additional learning is no longer worth its cost.

An organization cannot investigate every unexpected event. Doing so would create another form of waste, replacing premature correction with endless inquiry. What it needs is a way to distinguish consequential anomalies from irrelevant variation and turn those worth investigating into bounded decisions.

User's avatar

Continue reading this post for free, courtesy of Yetvart Artinyan.

Or purchase a paid subscription.
© 2026 Yetvart Artinyan · Privacy ∙ Terms ∙ Collection notice
Start your SubstackGet the app
Substack is the home for writing, podcasting, and media founders.