How Objective Are Data-Driven Decisions?

Today, data-driven decision-making is a religion.

We assume that decisions based on data are inherently objective.

But… is that true?

Consider, for example, if I gave you 30 years of Microsoft’s income statements and asked you: What should I expect next year’s gross margin to be?

Well, the formula is pretty simple. You divide gross profit by revenue, and you get the gross margin. So far, so good. Now you have 30 years of gross margin data.

But how do you decide what to expect for next year?

That requires determining which of the last 30 years next year will most resemble. For instance, should you focus on the last 12 months, the last 5 years, or the last 10 years? Should you exclude COVID-related years? The calculation itself doesn’t change, but the choice of timeframe does. The same data, when viewed through different lenses, can tell very different stories. It’s not just a numerical detail; it’s a contextual choice that shapes the narrative.

Does that mean data is subjective?

Data itself isn’t subjective, but what data to use is.

The selection of which data to analyze, how to analyze it, and even which questions to ask are all influenced by human perspectives, biases, and context. Data can reflect reality, but it doesn’t speak for itself—it’s shaped by the choices made along the way. The people collecting, organizing, and interpreting data bring their own biases, assumptions, and intentions, whether conscious or not.

Facts are always interpreted through a theoretical lens, and our theories shape which facts we focus on. The facts we select, in turn, influence our theories. This ongoing interplay suggests that data and theories are not separate; they evolve together in a cycle of refinement and understanding.

If data is subjective, what does that say about the decisions it drives?

Well, if data is subjective then it stands to reason that decisions derived from data are, to a degree, must also be subjective.

As data is filtered through choices, interpretations, and assumptions, the resulting decisions must also be influenced by these factors.

This doesn’t mean that data-driven decisions aren’t useful; it simply means they are not free from bias. They may be more transparent or structured than other forms of decision-making, but they’re still guided by human input.

Once we understand that data is not purely objective, we can use it more effectively to refine preexisting theories. The best approach is to start with a hypothesis and use data to invalidate it, rather than to simply confirm it. Begin with a working theory, make predictions based on that theory, and then see if the data aligns with those predictions. If it does, the theory might be in the right ballpark. If it doesn’t, the theory needs to be revised.

This is perhaps why analytics—which has been all the rage for years—has not delivered on its promise.

People expect analytics to deliver definitive answers from data. And that’s why they often get things backward. They begin with data, hoping it will reveal some objective truth, but all it does is confuse them.

Instead, they must start with questions and theories and use data to confirm or disconfirm those pre-existing theories. When approached this way, analytics stops being about data revealing absolute truths and shifts to iteratively improving theory that is closer to reality. And that is useful.

Ultimately, understanding that data-driven decisions are not purely objective—because data itself isn’t objective—doesn’t undermine their value; it just helps us use data more wisely. By using data wisely, we can refine our theories so they are closer to reflecting reality, which in turn improves decision-making.

If we don’t get that, we’ll keep drowning in data but thirsting for insights.

If you liked this article ?? Subscribe to Wisdom & Prosperity to join a community where we grow together. You will get an article delivered straight to your inbox every week.