Iron - Saude ao seu alcance

Tadeu Kwiatkowski Ribeiro

Question Discovery Champion

Tadeu Kwiatkowski Ribeiro is a data and analytics leader with over 10 years of experience designing, implementing and scaling solutions across healthcare, mining, travel and tourism and other sectors.

Tadeu combines technical expertise with business strategy and transformation, grounded in a Computer Engineering degree, an Executive MBA in Business Analytics and Big Data, continuing studies in strategic management and leadership at Stanford University, and hands-on experience building data and analytics capabilities in complex enterprise environments.

Rethinking Where Analytics Begins

One of the first lessons in data and analytics is simple: start with the business question. It is good advice. A dashboard built without a clear question can quickly become a collection of charts, and an analysis without a decision in mind can produce interesting numbers that nobody uses.

But that principle carries an assumption we rarely examine. What if the business does not know what to ask?

That question matters more today because analytics has advanced. Many organizations already have mature BI environments, detailed KPI systems, statistical models, machine learning, forecasting, segmentation, anomaly detection and increasingly AI-assisted analysis. Questions that once required days of manual work can often be answered much faster. The analytical toolbox is no longer the main limitation it once was.

The harder problem is knowing what deserves a question in the first place.

The Questions We Already Know

Most analytics work is naturally question-driven. Why did revenue fall? Which customers are leaving? Where are costs increasing? Which product is underperforming? What is causing a service-level problem?

These are important questions, and modern analytics can answer them with increasing depth. A team can move from descriptive reporting into diagnostic analysis, prediction and more advanced statistical methods. The problem is that all of this still starts inside the organization's existing field of view.

  The goal is not analytics without business questions. It is analytics that can generate the next business question.  

Requirements workshops make that visible. Sit with enough stakeholders, and you may collect dozens or even hundreds of business questions. Those questions are valuable, but they reflect what people already know, suspect, measure or consider important. The most valuable signal may sit outside that list.

From Answers to Discovery

There is another mode of analytics that deserves more attention. Instead of beginning only with "answer this question," it begins with a broader challenge: show me what deserves investigation.

That does not mean searching randomly through data until something looks interesting. It means creating room for exploratory analysis to reveal a structure that was not specified in advance.

A customer segment may be behaving differently from the categories the company normally uses. A small operational shift may appear months before it affects a headline KPI. Two variables that were never monitored together may reveal a pattern worth testing. A stable company-wide average may be hiding opposite movements inside different groups.

Techniques like exploratory data analysis, clustering, cohort analysis, anomaly detection, time-series analysis and statistical modeling can help surface those patterns. Their role is not to declare a new truth automatically. Their role is to create better hypotheses and better questions.

Discovery Still Needs Discipline

Open-ended discovery can also go wrong. With enough variables and enough combinations, it is always possible to find patterns that are accidental, misleading or statistically weak.

That is why discovery still depends on the fundamentals. The data has to be trustworthy. Definitions and grain have to be understood. The analyst needs a business context. Patterns need validation. Correlation cannot quietly become causation. A finding that looks surprising should survive scrutiny before it becomes a recommendation.

The goal is not analytics without business questions. It is analytics that can generate the next business question.

This also changes what we should expect from data teams. A strong analytics function should respond well when leadership asks a question. A mature one should occasionally walk into the room with something important that leadership did not ask for.

The Question Worth Asking

As BI, statistical methods, machine learning and analytical tooling become more widely available, simply having the ability to answer known questions becomes less differentiating. Competitive advantage may increasingly come from noticing important changes earlier, seeing relationships others have overlooked and challenging the assumptions that shaped the original questions.

The business question should remain the beginning of analytics, but it should not always be the boundary. A useful test for any analytics function is this: when was the last time the data team brought the business something important that nobody had asked them to look for?

Discovery does not replace the business question. Sometimes, its greatest value is creating the next one.