The Obsession with Being Data-Driven

This summer I was re-reading Sapiens, by Yuval Noah Harari. At one point in the book, he argues that there is an almost irresistible attraction toward the exact sciences; so much so that disciplines that are not — such as linguistics or psychology — dress themselves up as exact sciences to gain legitimacy, relying on mathematical formulas that often do not belong to them.

Drawing a parallel, I couldn’t help but imagine a pizzeria owner obsessed with finding answers as to why some of his customers don’t eat the cornicione (the crust of the pizza). Imagine he creates a database with the customer typology of those who leave the crust, how many pieces, whether it was lunch or dinner, how many people were there, and the average number of uneaten pieces… and he writes it down in an Excel spreadsheet after thousands of pizzas and empirical observations.

In the end, he has a spectacular dashboard telling him that 37% of his customers leave the crust, and that on Friday nights the figure increases by 18%. Any data-enamored mind would say: «Brilliant! We have the data». But the truth is that this number explains absolutely nothing about your pizza.

Wouldn’t it be infinitely easier to walk up to the table and ask: «Hey, excuse me, why didn’t you eat the crust?»?

Maybe they reply:

• «It’s just that the dough is chewy.»
• «For dinner, it feels too heavy for me.»
• «It’s too burnt.»
• «It seems dry compared to the center.»

Whatever the reason, all those behaviors that statistically represent the exact same data («didn’t eat the crust») have completely different solutions. Many product teams are like this pizzaiolo: they become obsessed with providing exact answers to problems that are not solved with statistical questions.

Maybe everything is much simpler than getting lost among Mixpanel, databases, and analytics dashboards. What if you sit down with your customers and talk to them? Perhaps you will discover their frictions in real time… and their stories.

Lumet’s Jury and the Right to Doubt

In Sidney Lumet’s wonderful movie 12 Angry Men, Henry Fonda challenges the opinion of the other eleven jury members, who right off the bat consider the accused guilty. When the others say «it’s unlikely», he maintains that innocence is possible.

The protagonist does not defend a closed hypothesis: he simply defends the right to reach the deepest truth, without absurd biases. Many organizations function like that jury; they do not lack data, but they lack personalities who question whether that data is sufficient.

The Voluntary Renunciation of Meaning (Sinnverzicht)

Unfortunately, for many, being data-driven means seeing only the quantitative side of the coin as the sole legitimate source. But for philosopher Hans Blumenberg, the problem goes far beyond a simple bias.

Blumenberg explains that, for daily life to function, we need stable routines. And that requires mechanisms that prevent us from constantly questioning the deep meaning of what we do, so we can focus on executing. If every time we made a transfer, made a medical decision, or passed social judgment we had to stop and reflect on the philosophy of justice, the value of life, or the origin of the universe, we would collapse under the weight of constant deliberation. The concept is Sinnverzicht: the voluntary renunciation of meaning. It is what allows us to pause the big questions to focus on the method.

Authors like Theodore Porter and Blumenberg himself point out three keys to understanding how this mechanism operates:

1. Rules as thinking machines. Algorithms remove mental fatigue. The mind relieves itself of moral deliberation if it simply follows rules blindly. We trade deep understanding for speed and efficiency: we learn to execute an action without needing to thoroughly understand how it works. Applied to our field: when a team sees that the churn rate rises to 4%, the automatic reaction is usually to launch a discount or add an aggressive pop-up to keep the metric in the green, without stopping to understand what real frustration led that person to cancel.
2. Coordinating at scale in a «society of strangers». In small communities, we decide based on mutual trust and face-to-face reputation. But in modern society, we coexist with strangers. To coordinate at scale, we need what Porter calls a «technology of distance»: standardized numbers and metrics that travel anywhere without depending on who looks at them.
3. The cost: decontextualization. For a piece of data to travel far and be understood equally by anyone, it must be flattened. If a teacher reduces all the effort and particularities of a student to a simple grade (an «A» or a «B»), they commit an act of Sinnverzicht. That letter is extremely convenient for averaging in a database, but it has destroyed the context and nuances of real learning.

The Illusion of Objectivity and Numbers as a Shield

Data does not speak for itself: it always requires interpretation. Quantifying at scale demands classifying reality into pre-established categories, and by doing this we decide in advance what we are going to remember and what we are going to forget. Often it is simply a facade of neutrality to evade responsibility.

Theodore Porter, in his book Trust in Numbers, argues that quantification is a way of «making decisions without appearing to be deciding». For bureaucrats and executives, it is ideal: by displaying work mathematically, they shield their decisions against criticism of arbitrariness, favoritism, or bias. If an official approves a grant because they believe in the project, they expose themselves to accusations of favoritism; if a formula with twenty indicators exists, it’s no longer them deciding, it’s the score deciding.

In digital products, the same thing happens: prioritizing a cold scoring (like a custom-inflated RICE) often only serves to avoid taking the risk of defending a qualitative bet before skeptical stakeholders. The number functions as a political shield.

There is a false premise that a mechanically objective method is, by definition, precise and truthful. A clear example is the 0.50 g/l blood alcohol limit: it is a mechanically objective method that anyone with a breathalyzer can replicate, but it does not accurately measure the true state of impairment or intoxication of each individual driver.

In science, a paradox occurs: communities with the highest prestige and internal cohesion — like high-energy physicists — rarely use rigid mechanical protocols in their daily routine; their decisions rely on expert judgment, mutual trust, and peer reputation. Conversely, in disciplines with fragile consensus or exposed to external scrutiny, rigid statistical rules are imposed as a shield to defend against suspicion and gain legitimacy.

Where does user experience (UX) fit in here? UX is a discipline with open and permeable borders that absorbs concepts from business, technology, and social sciences. Being subordinated to corporate priorities, it frequently resorts to the objective facade of metrics to claim professional legitimacy. The problem arises when we convert that human experience into flat data: we sacrifice the indispensable context needed to understand what the product really means to the person.

Renouncing the Search for the Cause of Things

A fundamental part of quantitative science decided at one point to renounce searching for the causes of things and settle for observing their correlations.

Why? The answer comes from late 19th-century positivism. For positivists, giving up on causes was a methodological victory for four reasons:

1. They saw causes as metaphysical concepts; in observable reality, there are only regular sequences of observable facts.
2. Because of the supposed neutrality of the observer: they believed that searching for deep causes projects personal biases.
3. To unify the method under the same mathematical tool applicable to any phenomenon.
4. To manage and control the world on a large scale with predictability. As Porter summarized: the world is not intrinsically statistical, but quantifiers made it statistical to be able to govern it.

Since an observational correlation does not prove causality, disciplines like medicine adopted randomized clinical trials to isolate variables. That is where the famous A/B test in digital environments originates.

The A/B test has its usefulness, but it carries a severe risk: placing a pre-checked box or hiding the unsubscribe button can cause the conversion rate to win indisputably in the short-term test, but at the cost of destroying user trust in the medium and long term. When the test becomes the sole arbiter of progress, we optimize the numerical representation instead of the actual quality of the product, incentivizing dark patterns.

The Perversion of Data-Driven

Many organizations have reduced being data-driven to a dangerous maxim: «everything that cannot be measured does not exist».

This does not only respond to technical inertia, but to a psychological and social need. On an individual level, the brain shuns uncertainty; seeing an indicator in green activates reward mechanisms and transmits an immediate — though illusory — sense of control. On a sociopolitical level, starting in the 1960s and 1970s, public trust in institutions and in the discretionary judgment of professionals eroded. Metrics were sold as a «technology of distrust». If a professional makes a decision based on their judgment and something fails, they expose themselves to punishment; if they shield themselves behind a numerical scale, they can always claim: «I didn’t decide anything; I applied the formula».

Historian Jerry Z. Muller perfectly describes this destructive spiral:

Initial lack of trust -> Demand for control metrics -> Professionals learn to game the system to protect themselves -> Cynicism and distrust grow -> And even more metrics are created.

The result is a culture that distrusts people’s judgment and replaces it with blind faith in dashboards and algorithms.

Toward Mature Data Governance

Mature governance does not seek to measure more, but to measure with criteria. To build it, I consider five pillars fundamental:

1. Data as a witness, never as a judge: Metrics should provide clues to enrich human judgment, never act as automatic sentences that impose decisions without deliberation or context.
2. Decouple high-impact incentives: Tying a metric directly to layoffs, bonuses, or promotions corrupts measurement; it incentivizes fudging and fraud. Mature organizations separate learning data from performance evaluations.
3. Co-design from the ground up: Indicators should not be imposed abstractly from an office; they must be designed in collaboration with the people who closely know the actual operations.
4. Multidimensional and adaptive metrics: Quantitative data must always be accompanied by qualitative stories, root cause research, and the direct voice of the user to avoid tunnel vision.
5. Protect «metric-free zones»: As the adage goes, not everything that can be counted counts, and not everything that counts can be counted. Human complexity does not fit entirely into a dashboard. It is essential to safeguard workspaces free from the pressure of a pre-fixed numerical goal to allow deep reflection and spontaneous collaboration.

Maturity consists of abandoning quantitative naivety and remembering that data should be at the service of learning and human judgment, not the other way around.

Epilogue: Reconstructing Meaning

Reversing this inertia is not easy because it requires rebuilding mutual trust within teams. Furthermore, the emergence of artificial intelligence adds a layer of complexity: if models limit themselves to processing quantitative directives at scale without ethical or contextual evaluation, they act as accelerators of Sinnverzicht (the renunciation of meaning) and Sinnverlust (the loss of meaning).

The solution does not consist of stacking more data or demonizing technology, but of recovering meaning: returning to question the purpose of what we build, remembering that automation lacks its own judgment. No dataset is entirely neutral; behind every algorithm there are prior decisions about what to register and what to discard.

The true challenge in product and design involves having the courage to question the inertia of numbers and restoring human, qualitative, and contextual judgment to the priority position it deserves.

If you’ve made it all the way here, thank you so much! If you enjoyed this piece, I’d love for you to subscribe for future posts. And if you’re into human-centered UX and digital products, check out my podcast, Pizza & Insights— it’s currently in Spanish, and I’d be thrilled to have you listen in!

Until the next post!

The Obsession with Being Data-Driven was originally published in Bootcamp on Medium, where people are continuing the conversation by highlighting and responding to this story.