As a data analyst, a significant chunk of my time is used to visualize data sets for different stakeholders. Data visualizations can take many forms. They can be a simple Google Sheets charts, some visualizations on a Python notebook, or a complex Tableau dashboard. Here are three questions that I always ask myself before visualizing any data set.
Who is your target audience?
This is undoubtedly the most important question to ask yourself. Are the stakeholders technical or non-technical audiences? What is their familiarity level with the data that you are about to visualize? Are they your peers or Executive-level managers? Are you visualizing data for a diverse group of audience or a specific team? Knowing your target audience well before thinking of visualizing data helps you achieve half of the success.
I often note down my audience profile by answering all of the questions above. I have different visualization approaches tailored to each audience type. For example, if my audience is a non-technical one who asked me to do some exploratory analyses, I’ll think of some Google Sheets charts instead of a Tableau dashboard. Doing that way, I can easily share with them the data set too. They can adjust the graphs or charts based on their needs. For example, they may realize later that they are only interested in last month’s data instead of the initial ask of 6-month data.
If I know that I am about to present some visualizations to a big and diverse group, I will pay more attention to adding more explanation into the visualization. For example, “daily active users” or “retention rate” may sound trivial at first. However, those metrics can have very different definitions across businesses or industries. I would add a short caption under my graphs or charts to explain how I calculated those metrics.
What is the objective of this visualization?
After knowing about my audience, I often initiate a quick discussion with them to learn more about their expectations. Simply asking “what is your objective?” may work, but it is not always as easy as it seems. However, I still recommend being very straightforward by asking about their needs and expectations. Then, I ask probing questions to boil down into the most important requirement that I need to meet.
For example, the Customer Success team wants to know more about customer complaints and their team performance. During the initial conversation, I would verify if they need an ad-hoc analysis or a deep-dive dashboard. In the first case, a few Excel charts should be enough to tackle their burning questions. In the second case, I need to build a dashboard showing various metrics to help them have a comprehensive understanding of historical data, current performance, and future trends.
Suppose that they want to have a dashboard. Then, we together brainstorm a list of metrics that they want to see. During the brainstorming step, I share with my stakeholder that they shouldn’t worry about feasibility or priority. Let’s just think of as many metrics as possible. After that, we rank the metrics based on priorities. If we know exactly what should be at the top of the priority list, this task is easy. Otherwise, we assign the relative scores for each metric, then choose the top three or five ones. Once the priorities are figured out, we talk about feasibility. Some metrics are easily built, some aren’t, mostly because relevant data used to build those metrics are not available yet.
The process that I mentioned might take one or two 30-min meetings. The list of ranked metrics is kept in a shared Google Docs file, which is sharable and allows others to comment. Having a clear structure about how you understand their expectations and requirements helps you navigate the data visualizing process much easier.
What is the story behind the data?
If the data doesn’t stand out to you, maybe you need to spend some time to think a bit deeper. For example, you might be very proud of a stacked bar graph showing the number of customer complaints broken down by categories. The graph looks clear with all the necessary headers and titles. The color palette looks great. However, what insights do you find by looking at this graph? You may realize that one category might dominate others. In this case, sorting the fields in descending order is the most helpful action. The audience can easily spot that dominating category in a first-place when looking at your graph.
Some types of data visualization look like the best candidate in the beginning, but not until you think about the story behind the data. When you are working with time-series data, it is very tempting to think that a line graph is a no-brainer answer. However, do you think sometimes a bar graph is a better candidate? A bar graph not only helps you see the trend, but it helps you spot outliers much faster, thanks to the outstanding strength of those bars compared to others.
Let’s go back to the Customer Success dashboard example. Suppose you are visualizing the number of customer service tickets through time with the x-axis being periods and the y-axis being the number of tickets. A line graph emphasizes your focus on the overall trend. However, a bar graph may switch your attention to a few weeks in August when there was a spike in the number of tickets. After doing some research, you may find out that the root cause was a new product launch during those weeks. Since customers were not familiar with the new product, they asked a lot of questions, which increased the number of tickets. As a result, the team realized that they should add more instructions about the new product to reduce the number of tickets for it.
Conclusion
These three questions are my self-made checklist before I do any data visualization project. That helps me navigate the process very effectively, saves time for everyone, and makes my workflow much more structured. My suggestion to you is to create your checklist. Please feel free to add more steps to the process if you find they are helpful. I hope you feel more confident when delivering any data visualization.
Originally published athttp://baovinhnguyen.comon October 16, 2020.
