Data Science Foundations Chapter 13: Telling the Story Behind Your Data

You did the hard work. You collected data, cleaned it, tested your models. And now you need to tell someone what you found. This is where Chapter 13 of “Data Science Foundations” by Stephen Mariadas and Ian Huke comes in. Communication. The part that separates useful data science from data science that nobody cares about.

I have seen this problem many times. Someone builds a brilliant model. Accurate predictions, solid math. Then they present it in a way that puts the room to sleep. The analysis dies on a slide deck that nobody opens again. Chapter 13 is basically a guide to not letting that happen.

From Data to Wisdom

The authors use the DIKW pyramid. Data, information, knowledge, wisdom. It is a classic model and it works well here.

At the bottom you have raw data. Process it, test your hypotheses, and you get information. Combine that with domain knowledge and real world context, you get knowledge. Communicate that knowledge well, and your audience gains wisdom. They make better decisions and take action.

Here’s the thing. Most data scientists stop at information. They show the numbers and assume everyone else will figure out what to do. That is not how it works. Your job is to get people to wisdom.

Think About User Experience

This might sound strange for a data science book. User experience? But the authors make a good point. Your report, your dashboard, your presentation is a product. And like any product, it needs to work well for the people using it.

They borrow the user experience honeycomb from Peter Morville. Seven principles: usable, useful, desirable, findable, accessible, credible, and valuable. Is your report easy to read? Does it answer the questions your audience actually has? Do people trust the numbers?

A practical example from the book. You could email a rainbow-colored Excel spreadsheet every Monday. Or you could build a live dashboard with one clear number that updates automatically. Same data. Completely different experience.

Know Who You Are Talking To

Not everyone needs the same information. The authors break audiences into four segments.

Technical audience wants the details. What model did you use? What were the parameters? Show them the math.

Managerial audience wants the key points in plain language. But they also want to be able to dig deeper into areas relevant to their team.

Executive audience wants the bottom line. Simple, clear, no jargon.

Team audience wants the key insights too, but maybe a more limited scope. Not everything is relevant to everyone.

The authors give a sales forecasting example. Executives want the overall number. Regional managers want their team breakdown. Sales reps do not need other regions’ data. And the data science team wants to know about the model. Same project, four different outputs.

One more thing. Be consistent across all these communications. If you tell the executives one thing and the managers something different, your credibility disappears fast.

Tell a Story, Not Just Numbers

This is probably the most important section in the chapter. The authors talk about data storytelling, which has its roots in data journalism.

Structure your findings like a story. They reference Gustav Freytag’s pyramid model, adapted for data: background, trigger, key findings, enlightenment, action. Set the scene, show what you found, explain what it means, recommend what to do.

But here is the key insight. Make the story about your audience, not about yourself. A data scientist might present contact center results as: “I took data from the source. I analyzed trends. NPS is declining.” Technical, accurate, boring.

Now change the subject to the customer: “Customer experience is falling. Their willingness to recommend us is dropping. If the trend continues, we will be in the bottom quartile compared to competitors within six months.” Same data, but now everyone in the room is paying attention. The stakes are clear. The story is about real people and real consequences.

Be Honest About What You Do Not Know

The authors warn about risks in storytelling. Every dataset contains many possible stories. You pick which one to tell. And that choice carries ethical weight.

Are you spinning the data to support a particular agenda? Are you hiding the stories people do not want to hear? In the contact center example, management might not want to hear about declining staff welfare. But that does not mean you should ignore it.

Also, correlation is not causation. Shorter call times correlate with lower customer satisfaction. But that does not prove one causes the other. Be upfront about uncertainty. Recommend A/B tests rather than presenting theories as facts.

Make It Last

If your findings have ongoing value, operationalize them. Build a repeatable process. Automate where you can. An ongoing report has different requirements than a one-time investigation, so revisit what stakeholders actually need.

Bottom Line

Communication is not an afterthought. It is the part that turns months of work into actual decisions. Know your audience. Tell a story that connects. Be honest about uncertainty. And please, stop sending rainbow Excel spreadsheets.

The authors’ practical tips are worth pinning to your monitor: understand your audience, plan the story, consider who the subject should be, and never tell people only what they want to hear.

Previous: Chapter 12: Model Evaluation Next: Chapter 14: Machine Learning and AI