How do you design data visualizations that are both beautiful and useful?

Best way to make data visualisation a super Hit, is to minimise the graphs, maximize the value by showing the most crucial KCIs and KPIs Secondly to use light colours, which is soothing to the end Users. Third, to minimise use of Pie charts or any fancy charts Fourth, just relax and let the user or focus group give his/her feedback for improvements if any.

I got great advice from my friend on this. Throughout the year, she jumps on calls with her work stakeholders, asking questions and learning how they use data, their goals and priorities. She also learns about the Key Performance Indicators (KPIs). Her approach is to keep in line with the company's priorities. This approach isn't just for employees. If you are looking for a job, you can do this too. Many companies make their KPIs public. Use this information to align your portfolio with what companies’ value most.

Knowing your audience is the most important aspect of data visualization. When I create data visualizations, I always follow these rules of thumb: #1 Use the right visualization for the data and the stakeholders. #2 Consider the level of detail that the stakeholders need. #3 Tell a story with your data. #4 And please keep it simple.

When it comes to useful data visualization, SIMPLE is the mantra. If it requires voice over or explanation, it's not a good visualization. That's the thumb rule I follow.

One thing I've found very useful is take your audience and your data both into consideration. You are basically trying to tell a story about your findings, don't overwhelm the audience by too much clutter. KEEP IT SIMPLE AND INTERACTIVE. By remembring this basic thing you'll be able to build a visualization which not only states facts and figures but also keeps the audience engaged and gives them a sense of contribution.

In the case of presenting visualizations, oftentimes simpler is better. The single most important concept and Analyst/Scientist must maintain is: How do I present this technical idea to a non-technical audience. While an intricate Tableau viz or advanced Seaborn plot will have it's place in certain circles, we should focus displaying complex information in the least complex way. Do not underestimate the power of a simple 2-D bar chart or line plot, especially when presenting information to senior leadership members!

Choosing the right type of visualization can be tricky, but a simple rule I have learned as data analyst is to create visualizations that make your data easy to understand and interesting to look at.

Choosing the right visualization involves considering your data's unique perspective, using color effectively, and recognizing its nature—categorical or numerical. Categorical data suits bar charts, pie charts, or stacked bars for comparisons, while numerical data thrives in line charts, scatter plots, and histograms, unveiling trends and relationships. Understanding this union ensures a seamless and impactful visual representation.

Imagine you have sales data with date, category of items, and market. You want to find the story behind the data but don’t know where to start. Using date and sales you easily produce a line chart to show trends at any time granularity. You notice a drop in sales in May and decide to investigate. Using the category of items column you break that line chart into a multi-line chart to see which category did poorly. You find out your baby clothes line is the product that was under performing. But with a line chart can you easily find out where? You create a table with a heat map to see all your data by market and quickly find out three markets are the problem. With the right data visualization you can quickly find the story in your data.

For proportions, I have found Pie charts to be unhelpful when the proportions are very close or when there are too many components. Especially when small differences make a big impact to the outcome. A stack bar is much better option in such cases. Pie charts however are great for showing significant change in proportions over time.

An example would be using the colour theme of your audience company. If the colours are flashy, it would be nice to use lighter shades/softer tones. Creativity is essential.

I have found Gesalt’s principles very helpful while designing visualizations. The Gestalt Principles, which include closeness, resemblance, continuity, closure, connection, and enclosure, are a set of principles that explain how the human brain processes visual information. There are basically 7 principles which can be applied in data visualization 1. Enclosure 2. Similarity 3. Continuity 4. Closure 5. Connection 6. Proximity 7. Symmetry

Here's a trick. If you add multiple colors to your visual, you should always include a legend. If the legend feels unnecessary - you dont need multiple colors. Color should add relevant context or information to a visual. For example, there is no need to give each category in a bar chart a separate color, becuase a bar chart already separates the data visually with the bars - the color doesn't add anything.

I love using color to increase intuition. For example, if I'm creating a dashboard and there are only a few categories I'm exploring, I like to create a color association for each category and keep it consistent throughout so it's faster to follow the information between various tiles. So Category1 would always be tied to ColorA, whether that's in a table with a conditionally formatted column or in a line chart or some other visualization. Category2 would be tied to ColorB in those same places. And similarly when I'm creating a presentation and posing a question, I'll highlight the answer in a visualization with a bright (but not too bright) color so it stands out as the obvious result, while leaving the other options more neutral.

I suggest to test the result also printing it in "black and white" to be sure that the contrast will correctly enhance any single color also in grayscale!

The worst thing you can do is pile as many visualizations as you can into a dashboard. When I was an Analytics Manager I would limit our dashboards to 3 main visualizations with any others needing to be done on different pages. Visualizations can quickly become overwhelming and useless if no one wants to look at them and actually use them!

You want your dashboard to inform as quickly as possible. Keeping it simple and declotteted will allow your user to get to the insights they need with minimal effort and will help to start building your reputation as a someone who can take data and knows what the user wants to see to take action.

I always follow the motto, “You aren’t done with your visualization when you can’t fit anything else onto the page, you are done when you can’t take anything else off.” Too often people ask the audience for what they want to see (the answer is ALWAYS everything) and they get stuck having to put too many “agreed upon requirements” in the report. What they really want is to answer their most important questions as quickly and efficiently as possible. Stick to the goal of the report. Try not to use more than 4-5 visuals on a page. Try to have at most 5-6 pages in the report, each with a specific purpose. As you are building the report, ask a trusted end user how they would use the report and take note of which parts they don’t need.

One of the most prevalent bad practices in data visualization are packed dashboards with lots of charts, filters, tables, icons, etc. You should overcome the need to "see everything". You should focus your analysis with the end user and the questions they are looking to answer in mind, as you probably heard, less is better. Focus on the quality of the analysis and easing the communication of the insights with clean visualizations.

It's helpful to have a basic template to ensure you have space for the most common/basic dashboard attributes. For example: a title bar, a filter pane, a key metrics section etc. Templates will force you to refine your ideas by setting boundries. Templates need to be living and iterative to stay relevant.

We use data to answer questions. Think about the problems your data might help to solve and possible questions that your audience might ask. This will help you select the right data and visuals to present. When presenting, use animations to guide your audience through your data, specially if your visual is complex to understand. By showing one data at a time, your audience will focus on you instead of trying to “decode” your visual

If you have to impress use transitions and animation, if you have to send a message and be supportive, better concentrate a coincise message on title and banner at the end of a slide. Consider that slides have to be e ported in PDF format to share with audience so avoid animation, transition and effect and concentrate in significative messages!

Report structure is extremely important. Typically I base my reports around how end users would chronologically want to see data. For example, page 1 compares products. Then they want to see a specific product so go to page 2 and filter to product and that page compares states. Page 3 may also filter to state and compare county. Etc. Having this setup where each page is useful and gets more granular is very intuitive for end users to use and will help guide them through their analysis very quickly and easily.

I like to use information panes or include a dashboard specs section as a standard build attribute. Heres what I normally include here: Dashboard description, Data Sources, Refresh Frequency, Last Refresh Date, Business owner, Developer name, Version info.

Test like a winner. 1. Compare against your design brief or success criteria. You should be solving a real business problem. 2. Validate that your metrics truly answer the question. Do consumers have to extrapolate information? 3. Sleep on it. If you come back fresh and struggle, your users will too. 4. Let your users show you how they think your visualization is used. Ideally, record the session, watch where they struggle and consider what guardrails are needed. Do you need more text or an icon to guide attention? Are your charts telling the audience what you think they are? Do the numbers mirror other sources of truth, and if not, can you find the reason (ex: a filter)? 5. Validate performance. Slow dashboards kill insights.

Don't forget to add a summary at the beginning of the visualization and divide it in subsection based on the summary linking the items to the right data visualization. Add also some key takeaways at the end!!!

In designing beautifully useful dashboards and visualizations Im often mindful of well first and foremost delineating - the baseline reporting, use cases, and audiences that derive value from the data, and how that value manifests. I like to set out from the beginning mindful of the value proposition to all stakeholders - and ensuring at bare minimum - key stakeholders are at least equally satisfied with the solutions delivering. One must be mindful of how data is used - is it manipulated / explored in pivots etc, is it used as further intermediates. Covering that baseline consider why comissioned - and the specifics and statistics whose story need to illuminate, and proceed with conduciveness to collaboration and corp okrs, kpis.

In the end, visualization is the graphic representation and presentation of data. You are showing your stakeholders what your data means in a compelling way using visuals. And as you build visualizations, consider your audience first and think inclusiveness and accessibility along the way.

Having beautiful reports is important. NOT AS IMPORTANT AS BEING USEFUL, but still important. If people like how your report looks and works, they will enjoy using it. The more they enjoy using it, the more they will actually use it, and one of the underlying goals of every single report is to have the end users actually use the report. Accurate results and answering an important question are the most important pieces when creating a report, but if you can hook your users into using your report with a good looking report, that can also be a big win.

To be artfully useful Always analyze visualizations everywhere around cross and keep a keen and enthusiastic eye out for data driven value delivery done better - as there are always doors open to continuous improvement, are artful analysts that architect value thats even superficially beautiful, and theres a rhyme and reason for every last decision point at the enterprise level. (Or at least should be). Often will be statistical analytics and studies distinguishing seemingly subtle yet significant differentiations. Using corporate colors and other cute approaches can sometimes be appropriate - but key thing is at end of day always rememember beaufiful visualizations are effective ones - tailored to doing their part to up kpis/okrs