The Completeness Fallacy: When More KPIs Lead to Less Understanding

Written by Andrei Costea on August 21, 2026. Posted in KPI

“The greatest danger in performance measurement is not seeing too little, but believing you’ve finally seen everything.”

Nine KPI Lessons, One Underlying Problem

Throughout this series, we have explored what might seem like nine separate problems with measuring performance. Each article examined a different symptom, manifestation, paradox, or unexpected consequence of using KPIs to understand a complex organization. Viewed on their own, these issues seemed only loosely related. When viewed collectively, however, they tell a surprisingly coherent story.

Our journey began with the KPI Theatre, where we observed how the act of measurement changes behaviour. Once individuals know they are being observed, they instinctively begin to focus on what is visible. This is not necessarily because they are engaged in manipulation or deceit, but simply because human attention is finite and people attend to that which attracts attention. Slowly, performance shifts from being good to looking good.

Then, we examined Goodhart’s Law to see what happens when that observation becomes a target.

What began as a useful indicator of performance becomes, over time, the objective itself. People stop asking whether the organization is serving its purpose and start asking if the numbers look OK. The simple act of optimizing has quietly replaced the process of understanding.

It would seem natural that if one metric blinds you, then two will be better, but that then begets something called KPI Saturation – when everything, every function, every initiative, every strategic priority, and every operational process is represented on a dashboard.

Everything becomes so overwhelming that we are blinded by what is actually present.

Information overload has given way to attention scarcity, and when attention becomes a scarce organizational resource, metrics themselves become valuable. In The Politics of KPIs, we examined how indicators gradually evolve beyond measurement tools into instruments of influence. As dashboards grow more central to an organization’s decision-making, knowledge that is difficult to reduce to a number becomes marginalized and increasingly ignored. 

Institutional knowledge, relationships, craftsmanship, intuition, judgment, and the trust required for collaboration – all fall off the strategic radar as information that cannot be easily expressed in numerical terms disappears from the conversation, as was illustrated by KPI Memory Loss. 

Organizations do not necessarily fail because they lack information. Rather, they fail because certain forms of knowledge no longer receive attention, or, even worse, are actively dismissed by those who treat them as irrelevant data points. The effects of these trends, however, do not remain confined to the organization itself. The final issues that we uncovered in this series impacted the very people who were employed to deliver on those numbers.

In the KPI Identity Trap, we witnessed how those being measured can become so completely identified with the metrics by which they are assessed that they forget to ask whether they are doing worthwhile work and simply focus on what the dashboard or scorecard shows. It could be seen that the last, and perhaps most fundamental point in this series – KPI Blind Spots – is almost a natural consequence. At best, a dashboard tells us something about the reality of our situation and, at worst, it tells us an unintentional lie.

Yet the lie is never told; it simply emanates from what is left unsaid: the part of reality deliberately omitted to focus on what is believed to be most important.

Realistically, none of these were separate problems at all: behaviour, targets, information overload, politics, organizational memory, personal identity, blind spots. These are not independent phenomena, but simply different facets of the same flawed logic. A logic that proceeds on the basis that reality, as it is currently expressed in KPIs, continues to surprise us only because we have not yet measured enough. 

This is what can be called the Completeness Fallacy, perhaps the most insidious belief underpinning modern performance measurement – the belief that performance (and measurement) can eventually be made complete.

Why Every Surprise Seems to Demand Another KPI

These are scenes every seasoned executive has lived out.

• An unforeseen event derails a critical business process. 
• A major customer segment starts leaving faster than anticipated. 
• A manufacturing defect escapes quality control. 
• A major project comes in late even though all milestone flags are green. 
• A cyber attack bypasses the security systems that should have prevented it. 
• A high-performing, tenured employee quits with little notice.

The post-mortem begins, with executives staring at dashboards and data visualizations, trying to pinpoint where the red flags should have popped up and when someone should have noticed that something was going terribly wrong.

One question invariably surfaces: “Didn’t we have a KPI for this?”

Sometimes the answer is yes, and it was simply ignored. Most of the time, more frequently than many organizations are comfortable admitting, the answer is no. Thus, the logical conclusion seems obvious.

“Let’s add one.”

On its face, this is completely reasonable. Every failure is an opportunity to refine the measurement system. If an important early warning was missing, then the dashboard should be augmented to track it. It often is the right solution, truth be told.

The danger emerges when adding a new KPI becomes the default response to unexpected events, because the next time the unexpected happens (and it always does), another KPI is added. Then another and another.

The dashboards get bigger, the reports get longer, the task manager bloats, and the analytical tools become more sophisticated. Still, unexpected events persist doggedly. Each surprise seems to reinforce the idea that something else must still be missing, driving the organization toward the impossible goal of complete measurement.

This is what can be called the Completeness Fallacy. 

The Completeness Fallacy is the mistaken belief that all organizational surprises stem from missing dashboard metrics and that simply adding enough KPIs will eliminate uncertainty entirely.

Complex organizations aren’t like simple machines, made of gears & cogs. They are living, breathing systems made up of people, incentives, cultures, relationships, informal networks, dynamic markets, shifting customer expectations, evolving technologies, and countless interactions that cannot be fully predicted. 

Every solution creates its very own new problems. Every intervention changes the system it attempts to measure. Having a complete representation of all possible futures is impossible. Ironically, as organizations pursue completeness, they move further away from true understanding because the question asked subtly changes.

We stop asking:

• “What have we misunderstood?”

Instead, we start asking:

• “What KPI are we missing?”

These two questions sound similar, but there is a profound difference. The first question is about understanding and assuming that the reality of the situation is more nuanced than the dashboard’s representation. The second question focuses on measurement and suggests that the dashboard just needs an additional piece.

The Endless Expansion of the Dashboard

Think about a dashboard that includes 50 carefully chosen KPIs. A few weeks pass, and then a completely unanticipated problem arises that those 50 KPIs couldn’t have foretold. 

“We need to add one more KPI!” – Leadership. 

The dashboard grows to 51 KPIs. A few months later, an even greater shock arises. A new KPI is added. The dashboard now has 60. Soon 80. Then 100. Eventually, someone gets tired and asks, “If we have 100+ KPIs on our dashboard, how did this still catch us out?“

It’s a strange psychological paradox at work here. 

• On the one hand, leaders think “everything that is important must be on the dashboard.” 
• On the other hand, when something isn’t on the dashboard, it is, at least initially, discounted precisely because it is unmeasured.

Whenever reality disappoints, we try to achieve completeness by expanding the dashboard. As the dashboard grows, our confidence in it grows. That, in turn, makes the next surprise that happens all the more baffling. People begin to wonder in dismay how all of this can be happening, since they are measuring everything.

Except they are not. They never can and will never be able to. No dashboard can perfectly mirror reality; reality is always larger than its reflection.

The danger isn’t what dashboards leave out; the danger is that we forget what they have to leave out.

Babies & Video Games: Why More Doesn’t Always Mean Better

The Puzzling Perplexity of Predicting Progeny 

The experience of dealing with babies is a universal (if not always enjoyable) one. Say your baby – a perfectly healthy baby, no less – is inconsolable, though not crying as a result of any obvious issue. They have been fed, cleaned, kept comfortable, and healthy, so why do the tears persist?

You hand the child a colorful toy, and for a brief moment, the wailing subsides before resurfacing with renewed vigor.

“Maybe the problem is simply that we don’t have the right toy!” and so you rush out and acquire one, only to have it achieve the same limited result. You then acquire another that sings, and another that flashes. The child continues to cry, and your resolve is steadfast: there has to be one “right” toy out there to soothe their little agitated spirits!

This process seems almost logical, and the conclusion (that another toy is just around the corner) feels almost automatic. After all, if there were a truly suitable toy, the baby would just stop crying. Right? RIGHT? 

This assumption misses a crucial detail: the baby wasn’t looking for another toy at all. Maybe they simply wanted to be held, or was bored lying in one position for too long, or maybe they wanted someone to talk to them, or just to feel the comfort of their parents’ closeness. The parents didn’t need a better toy; they needed to understand the baby. 

In the course of our lives, we will all learn an invaluable lesson. Sometimes the quickest, easiest path to resolution doesn’t lie in introducing a new component or finding a missing element, but in paying more attention to the subject of our concern.

Organizations are remarkably similar. Organizations often take a very different, and rather analogous, approach. Every unforeseen problem requires a new key performance indicator (KPI); every anomaly requires a new dashboard; every overlooked blind spot demands a new metric. 

It’s quite possible that the problem isn’t that another metric is needed. It might be that the organization needs to better understand what it is trying to achieve in the first place. Instead of reading reports, managers might need to speak with their employees more often. Instead of filling out surveys, customers might want to talk to real humans to air their frustrations.

For example, to understand a problem in production, supervisors should walk the factory floor rather than stare at a production-tracking dashboard. While the dashboard asks, “What else can we measure?” reality is asking, “Have you truly understood me?” That distinction is at the heart of the Completeness Fallacy.

The Curious Compulsion to Continue 

It plays out just as clearly in an area where organizational management can’t possibly be expected to surface: video games. 

Any person who plays role-playing games or massively multiplayer online games knows that optimization quickly becomes a way of life if left unchecked. Let’s assume you load one up, make a character, and it just isn’t putting out the damage-per-second (DPS) that you were aiming for. 

• The logical first step, you assume, is to get a DPS meter, take some measurements, and look for the source of the deficit. 

The meter tells you that you’re falling short of the damage output of everyone else on your team. You now have an answer to your problem. Or do you?

• Upon inspecting your equipment, you notice that some items are suboptimal, so you immediately replace them with better ones. 

You head back into the game’s content, only to see a minuscule difference in your damage output. There must be some other missing factor, something else you did not account for yet again.

• You realize your equipment isn’t enchanted, and so you spend hours painstakingly applying the most potent enchantments possible.

The result is still marginal. You now invest in better gems, talent points, food buffs, consumable potions, and specialization changes. Eventually, you end up with half your screen clogged by meters tracking DPS, timers, combat logs, raid frames, cooldowns, boss warnings, and who knows what else. Still, your character’s damage output does not noticeably improve.

Add-ons are useful, but the most obvious limiting factors, such as positioning, decision-making ability, encounter awareness, or knowing when not to attack, cannot be directly measured by another number or add-on. These qualities must be learned through trial and error and by recognizing patterns. However, the more information cluttering your screen, the easier it becomes to miss the giant boss looming directly in front of you. 

The Misguided Mission to Measure More 

Organizations can find themselves facing this type of music in exactly the same way. The dashboards continue to grow, the reports become more elaborate, the metrics increase, and the alerts pile up, giving executives an enormous volume of information to analyze. 

In practice, the actual amount of real-world understanding often advances at a far slower pace because it becomes so easy to confuse the process of measurement with actual understanding:

• You can measure customer satisfaction and learn that it’s trending downwards. However, you can’t measure the sound of disappointment in a customer’s voice on a support call. 
• You can check a productivity dashboard, which will tell you that the project velocity has decreased, but it won’t reveal the engineer who’s afraid to challenge a deadline they know is unrealistic. 
• You can monitor employee morale via an engagement survey, but that can’t capture the quiet moment in the hallway, months ago, when confidence eroded just a little bit more.
• You can assess sales conversion rates and identify where prospects drop off, but you can’t measure the trust that was lost in a single rushed conversation. 

Wisdom doesn’t grow automatically simply from an increasing quantity of observations. Wisdom is built through an interpretation of those observations within the human context in which they occurred. 

This is the true danger of the Completeness Fallacy: it persuades leaders to believe that the missing piece of the puzzle is simply another data point, and that there is never a need to stop looking at data and start listening to their people.

From Hospitals to Software: How Industries Share the Completeness Fallacy

The form in which the Completeness Fallacy shows up may vary significantly between industries, but the core principle doesn’t: whenever the world spits out an answer that surprises us, we start hunting around for some new measure, instead of asking a question: “Are we perhaps measuring the wrong thing to begin with?“

Healthcare: Measuring Patients While Missing Care

Modern hospitals amass vast quantities of data: Patient wait times, Bed capacity, Readmission rates, Length of stay, Medication adherence, ED wait times, Length of procedure, and Infection rates, among many others.

They’re all crucial metrics, but none of them fully explains why some patients with similar clinical profiles take weeks to recover while others take days. Many of the things that drive outcomes – whether the patient actually grasps what the doctor just said, has someone to nudge them about medications, or simply trusts their physician with what’s bugging them – are tough to fit into a standard spreadsheet. 

We see a bad readmission outcome and naturally reach for a new quality measure. Sometimes that makes sense, but more often than not, we need to make sure the conversation was handled properly, not simply as a way to create another quality benchmark.

Manufacturing: Perfect Machines, Imperfect Systems

Manufacturing organizations usually have extremely detailed operational dashboards: Machine usage, Cycle time, Defect rates, Overall Equipment Effectiveness (OEE), Scrap rates, Downtime, or Energy usage, to name just a few. 

When product quality drops inexplicably, the first reflex is often to monitor it even more closely. Very often, the mentality is “time to add another production KPI and slot in another quality check.”

Yet, investigations often find that technical issues are not at the heart of the matter; rather, the heart itself is gone.

• An experienced operator retired. 
• Shift handovers became hurried. 
• Maintenance personnel stopped communicating informally. 
• Production targets subtly discourage the workforce from signaling minor deviations until they become significant ones.

None of these had anything to do with a lack of data regarding the machinery – these were all human systems impacting technical ones. What was needed wasn’t more sensors but a better understanding of those humans operating them.

Software Development: Measuring Productivity Without Seeing Complexity

Software teams today arguably measure more things than any team has before: Sprint velocity, Story points completed, Lead time, Cycle time, Deployment frequency, PR approvals, Bug counts, Code coverage, or Incident response time.

When the wheels are slowing unexpectedly, leadership frequently adds a new metric for engineering teams to work with. 

• “Maybe we need more code!”

• “Maybe the reviews are taking too long!”

• “Maybe the deployments aren’t happening often enough!”

Maybe the real problem might be that a legacy architecture is now extremely difficult to maintain. The team has quietly accumulated technical debt over the years and is now spending much more time understanding systems than developing features, and on top of that, cross-functional communication has broken down.

The biggest thing stopping the team from shipping may not be an absent KPI, but decades of accrued complexity that only a truly seasoned engineer can even see. Such nuance cannot be encapsulated in a dashboard.

Aviation and High-Reliability Organizations: When Safety Lives Between the Metrics

Few industries are more serious about measurement than aviation. Aircraft systems generate vast amounts of operating data, which are measured and remeasured with exceptional rigor. 

• Flight schedules are closely monitored.
• Safety incidents are rigorously documented.
• Procedures are formalized. 
• Performance is routinely assessed. 

Nevertheless, aviation professionals know a secret that few leaders recognize: if no safety event has occurred, it doesn’t mean that safety is present.

The organization could show perfectly good safety performance on its metrics, while easily and effectively hiding its own psychological limitations regarding the issue from itself. Pilots may be unwilling to speak about hazards. Maintenance crews might be reluctant to acknowledge near-miss events because they might feel embarrassed or appear incompetent. A Junior team member might have no idea how the process could be made safe and would have no incentive to raise such concerns.

The overall safety indicators might be beautifully and blissfully green right up to the day of a disaster. When such a disaster occurs, the investigators rarely attribute the failure to a single additional KPI or a particular missing index or tool. Instead, they generally attribute the disaster to communication failures or human errors that may have predated the measurement systems.

The organization didn’t lack yet another metric – it lacked a deeper understanding of the system it thought it had measured.

In all 4 of the industries we’ve showcased, the situation repeated itself: 

• Disasters occurred
  • Organizations assumed they must be missing a metric
    • The metrics were expanded
      • They were surprised again and again

Such scenarios don’t happen because leaders lack the intelligence to be able to design and implement appropriate measurements, but because no complex system, as a matter of principle, can ever be as large as its own representations of itself. No dashboard can become the entity being measured.

When organizations reach that stage, they focus so intently on measurement devices that they neglect the journey and miss where they are going.

Final Thoughts

Looking Through the Windshield Instead of at the Dashboard 

Never before has there been so much readily accessible operational information available. Never before has there been a way to see performance across continents in real-time, spot emerging trends within minutes, or turn vast amounts of raw data into simple, intuitive pictures. Organizations have become faster, more coordinated, better informed, and ultimately, more agile. 

Problems arise only when dashboards stop being tools for understanding and start becoming substitutes for it. This is, ultimately, what’s dubbed the Completeness Fallacy. 

It has nothing to do with whether KPIs are useful or useless, or whether measurement systems need upkeep or not. Rather, it is the false belief that uncertainty can eventually be engineered out of the organization by judiciously selecting enough indicators.

In modern-day organizations, every KPI answers one question but begs another. Every bit of extra visibility alters behaviour, thereby changing what the measurement actually means. Performance measurement isn’t a race towards a state of completeness. It’s a never-ending conversation between what can be measured and what requires only observation, judgment, curiosity, and experience.

Dashboards are the thing that should inform a conversation, never replace it. Think of them as a car windshield. It has never been designed to capture every detail of the road ahead. It won’t reveal what is hiding behind every building, nor will it show every hidden pitfall lurking around the bend. Its purpose is to provide just enough visibility to help us navigate while, at the same time, reminding us that the road itself deserves our attention.

Modern performance dashboards have precisely the same role. They are not the organization, but the windows through which we may view organizations in a richer, more dynamic manner than any collection of KPIs can adequately convey.

If we start to look at the windshield instead of through it, we are liable to lose sight of where we are going, and that may well be the most profound irony of modern performance management:

We have never measured more, and with better tools. At the same time, we have never depended more on the interpretation, conversation, experience, and context provided by the people powering the soul of our workplaces.

*********

Looking to strengthen your expertise in KPI design and performance measurement? Gain practical knowledge, proven methodologies, and globally recognized certification through The KPI Institute’s Online Certified KPI Professional.

Tags: Completeness Fallacy, dashboards, Decision making, Goodhart's Law, information overload, KPI, KPI Saturation, Organizational Performance, Performance Management, Performance Measurement