Decisions without data: how to build a data-driven culture and stop making gut-feeling choices

In 20 years of strategic consulting, I have heard the same phrase in meeting rooms at banks, fintechs, retailers, and startups: "in my experience, I think we should follow this path". Experience has value. The problem is when it replaces data instead of complementing it. The difference between companies that grow sustainably and those that keep putting out fires lies, in large part, in this distinction.

Building a data-driven culture is not about buying a BI tool or hiring a data scientist. It is an organizational mindset shift that starts at the top and permeates every decision, from operational to strategic. And in Brazil, despite advances in digital transformation, we are still a country of gut feelings disguised as strategy.

The real cost of guesswork in organizations

Before talking about how to build a data-driven culture, we must be honest about what it costs not to have one. I am not talking about abstract costs. I am talking about money, time, and lost opportunities that show up on the balance sheet — just in lines that no one can clearly identify.

An IBM study estimated that poor data quality costs American companies around $3.1 trillion per year. In Brazil, the scenario is proportionally similar. When a company launches a product based on the sales team's perception instead of customer behavior data, when a CTO decides to migrate the entire infrastructure without performance benchmarks, or when a CEO sets growth targets based on a market "gut feeling," the risk is not hypothetical — it is concrete.

I worked with a large Brazilian financial institution that made a regional expansion decision based on the enthusiasm of three executives who "knew that market well." Two years later, the numbers showed what the available data had already indicated: the target market had a default risk profile incompatible with the product being offered. The loss amounted to tens of millions. The data was available. It just was not consulted.

What it really means to be data-driven

There is a dangerous misconception in the market: many companies think they are data-driven because they have dashboards. Having a dashboard is not being data-driven. It is having a dashboard.

Genuine data-driven management has three fundamental dimensions:

• Data as a decision input, not as opinion validation. The difference is subtle but devastating. In the first case, you collect data, analyze it, and decide. In the second, you have already decided and look for data to confirm your view. The second behavior is called confirmation bias, and it is endemic in organizations.
• Democratization of access to information. When only the analytics team has access to data, you create a bottleneck that paralyzes decision-making speed. Product leaders, operations managers, and even customer service teams need access to data relevant to their context.
• Accountability for measurable results. Data-driven decisions need success metrics defined before execution, not after. This creates real accountability and feeds a virtuous learning cycle.

When I worked at AWS helping Brazilian companies structure their generative AI strategies, the biggest obstacle was rarely technical. It was cultural. Companies wanted ready-made models but did not have data governance to feed them with quality. Garbage in, garbage out — it applies to AI and to any business analytics.

The four pillars of an effective data strategy

A robust data strategy does not emerge from nothing and cannot be solved with technology alone. It requires four pillars working in an integrated way:

1. Data governance

Who owns each piece of data? Who can access what? How do we ensure quality, consistency, and compliance? Without governance, you have a repository of unreliable data — and unreliable data is worse than having no data at all, because it creates false confidence. Governance does not need to be bureaucratic, but it needs to exist.

2. Scalable infrastructure

It is not possible to build a data-driven culture on fragile infrastructure. Companies that grow quickly frequently face the problem of data distributed in silos: one system for finance, another for CRM, another for operations, none of them talking to each other. Modernizing this infrastructure — with data lakes, data pipelines, and cloud architectures — is what enables real-time and large-scale analysis.

3. Data literacy and capacity building

One of the biggest mistakes is outsourcing the responsibility of interpreting data exclusively to analysts. CEOs, CTOs, and managers need sufficient data literacy to ask the right questions, challenge analyses, and interpret results with critical thinking. This does not mean knowing how to code in Python — it means understanding what a piece of data represents, its limitations, and its context.

4. Structured decision-making processes

Having data available is useless if the company's decision-making process does not include consulting it. Strategic meetings need to have data as a mandatory agenda item. Investment approvals need to have quantitative analyses as a requirement. This seems obvious, but in practice, most companies still make decisions in meetings where the most important slide is the opinion of the most senior executive in the room.

How to start the transformation in practice

The question CEOs and CTOs ask me most frequently is: "Where do we start?". The honest answer is: it depends on the company's maturity level. But there is a path that works regardless of the starting point.

The first step is always a data maturity assessment. Map where the company's data resides, what its quality is, who accesses it, and how critical decisions are made today. This assessment tends to reveal uncomfortable surprises — and that is precisely why it is necessary.

The second step is to identify two or three high-impact use cases where data can make an immediate difference. Do not try to boil the ocean. Choose real problems: churn reduction, inventory optimization, conversion improvement, default management. Solve these problems with data, show measurable results, and use these cases as internal proof that the approach works.

The third step is to create executive visibility. Data that does not reach the C-level does not change culture. A simple executive dashboard, updated in real time, with the metrics that truly matter to the business, is more powerful than any PDF report sent weekly.

The data-driven transformation does not begin with technology. It begins when the CEO stops a meeting and says: "Before we decide, what does the data show us?"

Pitfalls that sabotage data-driven culture

I know companies that have invested millions in analytics platforms and still make decisions based on gut feelings. There are recurring patterns that explain this failure:

• Buying technology without changing processes. A tool without a process is a waste. Before hiring any platform, define which decisions will be made based on data and how that process will work in practice.
• Creating an isolated data department. When analytics is a silo, other teams do not feel responsible for the data they generate. A data-driven culture requires each area to be a co-owner of the quality and use of its information.
• Chasing perfection before starting. Many companies remain paralyzed waiting to have data that is "clean enough" to begin. But data will never be perfect. Start with what you have, improve iteratively, and generate value while evolving quality.
• Not connecting data to business outcomes. Analyses that do not answer concrete business questions do not change behavior. Every data project needs to start with the question: "Which decision will we make better with this?"
• Leadership that requests data but does not use it. Nothing sabotages a data-driven initiative more quickly than a leader who requests analyses and then ignores the results when they contradict their intuition. The signal this sends to the organization is unequivocal: data is decoration.

The role of generative AI in the data-driven evolution

I cannot talk about data-driven decisions in 2025 without addressing the impact of generative artificial intelligence on this equation. Generative AI is democratizing access to sophisticated analyses in a way that was unthinkable three years ago.

Today, tools with natural language interfaces allow an executive to ask complex questions about data without knowing SQL or programming. "Which regions experienced margin decline last quarter and what is the correlation with sales seasonality?" — this type of analysis, which previously took days with an analyst as an intermediary, can now be done in seconds.

But beware: generative AI amplifies both the good and the bad in your data strategy. If the database is of poor quality, the AI will generate incorrect analyses with an appearance of sophistication. If governance is weak, the AI will expose sensitive data in ways that data protection regulations will not forgive. The companies that will reap the greatest benefits from AI applied to data are precisely those that have already built a solid foundation of governance, quality, and culture.

Companies like BTG, XP, and Inter — with whom I had the opportunity to work — did not reach the forefront of AI by accident. They got there because they spent years building data infrastructure, governance, and culture before placing algorithms on top. AI was the accelerator, not the starting point.

The decision that changes everything

Building a data-driven culture is, fundamentally, a leadership decision. Not a technology decision, not a budget decision, not a hiring decision. It is a decision about how the company will operate, about what will be valued, about which behaviors will be reinforced and which will not be tolerated.

Companies that have successfully made this transition share one characteristic: leadership was the first to change. The CEO was the first to ask for data before deciding. The CTO was the first to require metrics before approving projects. When the top of the pyramid changes, the rest follows — not immediately, not without resistance, but it follows.

The Brazilian market is at an inflection point. The combination of greater data availability, falling cloud infrastructure costs, and the maturity of analytics and AI tools has created a window of opportunity that will not remain open indefinitely. Companies that build a solid foundation of data-driven decisions now will have a structural competitive advantage over those that continue operating on gut feelings.

The question is not whether your company will become data-driven. The question is whether it will do so before or after your competitors.

If you are at a decision point about how to structure your company's data strategy — whether starting from scratch or evolving an existing initiative — I can help chart a realistic path, based on your business context. Get in touch at abraao.tech and let's talk about how to turn data into real competitive advantage.