The goal of a data professional is to present correct and actionable insights. The transformation from data artist to data scientist begins not with code or math, but with the words we use to describe our findings.
In my military service, I had to review intelligence reports with junior analysts. These sessions were exhaustingly long and held in dull offices, but had tremendous impact. It was almost Kafkaesque. The analysts would try to distill all they learned about a subject into a digestible, high-skim-value executive report, while I had to make sure every sentence was correct and actionable.
“We assume the threat is credible,” one of the analysts read from the report and already started reading the next line.
I interrupted him.
“Do we?” I asked. “Or did someone study it?”
That one word—assume—could be the difference between informed decision-making and dangerous oversimplification.
Language is the logic map you hand to others so they can follow your reasoning, or challenge it. We must not treat language like it is flexible or decorative. If our job is to draw meaning from data, then words are our models. And like any model, they carry structure, confidence, and risk.
This column is about why lingo matters and how your choice of words reflects your reasoning, your evidence, and your integrity. You will come away with a framework you can apply in your reports, presentations, and peer reviews, so you can be precise, readable, and responsible all at once.
Language as Methodology
When we are working with data, our conclusions are not built solely from numbers; they are built from the interpretation of those numbers. And that interpretation is carried entirely by our language. This is why we should treat language the same way we treat model architecture: deliberately.
There is a world of difference between saying I assume, I infer, or I assess. Each of those signals a different degree of confidence, evidence, and methodological rigor.
For example, calling something an assumption signals that no one has verified it, and could imply it was not even examined. But if someone has studied it, then assessment is the correct term, as it implies a formal process: data collection and analysis, logic, and accountability.
This kind of linguistic precision enables honest and confident decision-making by exposing the scaffolding behind your work. It gives your reader a map of your reasoning for them to judge its strength.
The Knowledge Pyramid: A Model for Analytical Thinking
To explain the flow of logic from raw data to structured thinking, I often reference a variant of the Knowledge Pyramid [i], [ii]:
It is a deceptively simple hierarchy, but essential to how we think and speak.
At the bottom is data: objective and trusted up to the limitations of its source and sensor. Above that sits information (also called “insights”): the first level of interpretation. That’s where patterns begin to form and meaning starts to emerge. Then comes knowledge: validated and generalized insights. And at the top is wisdom: the judgment to act effectively, grounded in layered understanding and context.
Here’s the catch: Most of what we present as analysts live above the data layer. We rarely deliver raw logs, but rather interpreted, sometimes abstracted meaning. And in that gap, language does the heavy lifting.
We lean on constructs like assumptions, beliefs, inferences, projections, and axioms, each of which represents a different kind of logical structure. Some are philosophical; some are statistical; some are just habits. But all of them shape the bridge from “what the data shows” to “what we recommend.” If we don’t label those layers with care, the entire structure risks collapse.
Conclusion: Clarity, Responsibility, and the Analyst’s Edge
As data scientists, our audiences can usually only judge our final product. We choose what to model, how to frame it, and what to report. This grants us immense power, but with it comes a great responsibility: expressing not just what we know, but how well we know it.
So, start small. The next time you write a report or present findings, ask yourself:
- Am I openly stating my assumptions?
- Am I using higher-tier language faithfully?
- Am I showing the scaffolding of reason and logic beneath my conclusions?
And when you read others’ work, read it actively. Look for their language cues. Consider how they got to their conclusions, and what’s missing from their logical map?
Because the transformation from data artist to data scientist doesn’t happen in a Python notebook. It happens in the words you choose, and the ones you refuse to let slide.
[i] Wikipedia contributors. (2025, May 25). DIKW pyramid. In Wikipedia. Retrieved October 3, 2025, from https://en.wikipedia.org/wiki/DIKW_pyramid.
[ii] Lucky, R. W. (1989). Silicon dreams: Information, man, and machine. St. Martin’s Press.
Columnist: Yoav Goldhorn