An experienced data scientist knows how to tell the story of her data effectively. To make the main points of her data visualization stand out, and the interesting parts catch the eye of the audience. Throughout the multitude of presentations a data scientist does during her career, she learns quite a few lessons about the making of powerful data stories and diagrams. I surfed the web to find some of the best parts of this advice for you.
- Use high-quality data. You can never make gold from copper. That’s why the pursuit of alchemy is dead, and why you mustn’t take your time trying to present insubstantial results you got from low-value data. This really doesn’t work. At least not where you want it to work.
- Know thy data (and audience, questions, and goals). You can’t show others what you yourself are unable to see. To make good visualizations, you must have thoroughly explored your data; find and understand the patterns within your records, and take note of why is that important to your audience.
- Choose the right diagram. Some plot types are known to all. Most of you have draws Bar plots, Pie charts, Line plots, Scatter plots, histograms, maybe even Heat maps, Treemaps, Word clouds, and network diagrams while at school or work. Even so, there are variants of these graphs you’re not familiar with, and sometimes you learn that they can be used with unexpected kinds of data. The others are not so.
- Keep it simple. Leonardo Da Vinci says: “Simplicity is the ultimate sophistication”. If you can’t make your point in a non-complex way, you have not understood it well enough. Bring the necessary points to your reader’s view. Don’t overcrowd the charts with unuseful details and ornaments.
- You still need to write some things: Write artistically. Mind your labels, titles, legends, axes. They should be no longer than is necessary, but never so short that they lose their expressivity. Never miss a label where needed. Never repeat a label twice (e.g. bring bar labels I both in the legend and beside the bars). This takes away the simplicity of your work.
- Order and Format the layout of your report so that it can be understood at the first glance. Order the elements in your visualization consistently, intuitively, and evenly. Your audience shouldn’t have to think to find the order of the items in your presentations. Like any other side detail, the order and format of your diagram must be so obvious that they don’t divert the viewer’s attention from the point you’re trying to express.
- Never ignore the effect of coloring. Colors bring life to your visualization. Besides that, it adds another dimension to your diagrams where you can express new information. Heatmaps are a fine example of how proper coloring can help you understand a ton of data in a glance.
- Accentuate your work by changing the size of its elements. Word clouds are fantastic charts: they beautifully present you with the big picture of a large dataset in an instant, without using any numbers. You can differentiate the size of the elements of your visualization to emphasize the more important parts or to compare the magnitude of a feature between different entities.
- Add different shapes to the items in your visualization. The goal of statistical visualization is to minimize the amount of reading your audience needs to understand your point. Replacing words with creative, cute shapes make your diagrams attractive and enjoyable. At the same time, it makes your report easier for the eye.
All this said, using too many ornaments is one of the worst mistakes in a presentation. Never do it if your work is already crowded with different shapes and colors, or if you cannot find minimally-designed, distinguishable icons.
- Your charts should tell a story. If your goal is to catch the eyes and minds of your audience, you must make a strong story out of your findings. Thoughtfully plan how are you going to present your work. What are the titles? In what order the charts are displayed? How are you going to move the attention of your readers/listeners from one part to the other? How are different sections in your presentation are laid out? What is the bigger picture you want your audience to perceive? You must have good answers to all these questions.