What Are the Most Important Elements of Data Storytelling?
Data scientists are at their most effective when the insights they share are clear and actionable. How to get there will depend on the quirks of the data you have and the stakeholders you communicate with, but there are a few key areas where it (almost) always pays off to invest our attention. Here are three recent posts that focus on specific elements of compelling storytelling—and do it extremely well.
- Don’t let your dashboards gather dust. A lot of work goes into building data dashboards, but far too often, the resulting product is confusing, clunky, and underused. Marie Lefevre believes effective and useful dashboards are within reach. She shares a four-step framework—based on personal experience—that will help you build dashboards that tell data stories with precision and clarity.
- Finding the right framing is key. As an educator, Alejandro Rodríguez knows that even the most complex concepts can become approachable if you choose your communication method wisely. This post is ostensibly an introduction to confusion matrices and classification metrics, but it’s also a masterclass on the power of a simple, well-chosen example. (In this case: poison mushrooms!)
- The crucial importance of colors. Data visualization is about making massive amounts of information accessible and interpretable. Its success depends on a series of design decisions, both small and big; Weronika Gawarska-Tywonek’s excellent primer will help you understand how color palettes work, and how to go about choosing the one that’s most appropriate for the task at hand.
Our other publishing highlights this week go far and wide beyond data storytelling—whether you’re a job seeker, a seasoned ML practitioner, or a data-for-change advocate, we think you’ll find something here that’s well worth your time.
- How do you manage the configurations of your model training experiments? Suneeta Mall’s deep dive covers Pydantic and Hydra and the ways these tools might simplify your configuration workflows.
- If you have a data science job interview coming up (or even if you don’t), check out Tessa Xie’s overview of the distribution-related concepts you need to be fluent in.
- “Interest in using simulations to generate massive amounts of synthetic training data is steadily growing,” says Mason McGough, who goes on to show how you can generate 3D models of human beings for computer vision.
- The quest to minimize AI’s carbon footprint is already underway; David Yastremsky’s overview explains where we need to focus to make deep learning projects truly climate-friendly.
- If you’re curious about dynamic programming but have yet to find a patient, thorough guide to show you the ropes, Han Qi is here to the rescue with a step-by-step explanation of the Cherry Pickup problem.
We hope you enjoyed learning and exploring on TDS this week! If you’d like to support our authors’ work, consider becoming a Medium member (which also makes our entire archive available to you).
Until the next Variable,
TDS Editors