Introduction
Most data visualization courses and certification programs do an excellent job of teaching students the technical skills needed to build data visualizations. However, they don't reflect all of the skills needed to conduct a successful data visualization project for real businesses.
To be precise, programs don't put enough emphasis on mastering the non-technical side of data visualization, which I believe is one of their greatest limitations. From my past 4 years creating and presenting visualization tools, I've come to learn that visualization projects require more time to be allocated to the non-technical elements than the technical elements.
Furthermore, mistakes that stem from poor technical skills are relatively easier to rectify, whereas mistakes that stem from poor non-technical skills tend to create significant setbacks.
Courses may touch upon the importance of soft skills like communication and leadership in data visualization, but they don't provide much insight into how to harness or develop those skills. For instance, I've seen the advice "know your audience" get preached a lot, but such vague advice isn't really actionable.
To address this major shortcoming, I've drawn on my past 4 years of experience to compile 6 actionable tips for executing data visualization projects successfully.
Tip 1 - Define the Scope
Oftentimes, clients or stakeholders make vague requests for visualization tools. They are either too busy to detail the specifications or simply don't know what they exactly want in the first place.
For successful data visualization, analysts should first define the scope of the project by asking clarifying questions.
Asking questions not only sheds light on what should be included in the visualization but also indicates what should not be included. This saves data scientists/analysts time while also preventing them from delving into topics that the clients aren't even interested in.
The most relevant questions to ask depend on the scenario, but commonly asked questions include the following:
1. Why is this visualization needed? Will it be used for monitoring purposes or will it be used to make business decisions?
2. What are the key metrics?
3. Who are the end users? How will they access the visualization?
4. Will the visualization be updated? If so, how often?
Tip 2 - Understand the Data
It's worth allocating some time to become familiar with the underlying data used to create the visualization.
Analysts need to understand the data they are working with so that they can contextualize the numbers reported in the visualization.
This might seem like an unnecessary step, especially if you're accustomed to working with the toy datasets provided in many visualization courses. However, datasets provided by clients can comprise hundreds or even thousands of columns!
Creating visual charts without a sufficient understanding of the data will increase the likelihood of an analyst reporting erroneous values without even realizing it.
To avoid this outcome, analysts can examine the available data by:
1. Manually skimming over the data
2. Writing SQL queries
3. Using data analysis libraries in Python and R.
Tip 3 - Conduct Sanity Checks
Even after a team builds a visualization with close attention to detail, mistakes are bound to be made.
To prevent any egregious mistakes from reaching the eyes of the clients, the visualization tool must be subject to auditing. You can conduct the sanity checks yourself or leave them to other members.
When testing the dashboard, take on the perspective of a user and consider the following:
1. Does the visualization adhere to the client's specifications?
2. Does the relevant KPI stand out in the visualization?
3. Do the charts have appropriate titles, axes titles, and legends?
4. Do the reported numbers in the visualization make sense?
5. Do the design elements in the visualization (e.g., color, spacing) make the contents easy to digest?
6. Are the interactive features in the tool (if any) intuitive and easy to use?
Tip 4 - Solicit Feedback from Clients
Soliciting feedback from clients is a great way to assess the quality and effectiveness of the visualization tool. It gives clients the opportunity to monitor the progress of the data visualization project and provide direction for the subsequent steps.
In many cases, the specifications for a project detailed in the beginning are tentative and may not even be feasible. So, showing clients your work gives them the chance to change their requirements as needed.
The clients' input is also needed to verify that the visualization is appealing. Since the tool is made for the clients, they will need to provide their opinion on the more subjective elements of the visualization. Furthermore, when building an interactive visualization, data scientists/analysts need to give clients some time to play with it themselves so that they're comfortable with using it.
Tip 5 - Avoid Being the Single Point of Failure
A single point of failure is a person whose absence would render their team unable to function.
While it's ok for a team member to be solely responsible for certain duties in a project, they should not be irreplaceable; other members should be able to step in and perform the role adequately in their absence.
For long-term visualization projects, you'll likely be unavailable to help in certain stretches. A part of your responsibility as a team member is to ensure that others can take over your role when necessary.
For instance, if one of your duties was to update a dashboard with new data periodically after its deployment, you would have to make sure that other members of the team are able to make the updates themselves when you are not present.
There are two primary ways to prepare team members to take on your responsibilities.
- Train team members
When you identify yourself as a single point of failure, you can arrange meetings or calls to prepare others to take on your role. Training the team is a convenient option as it can be done on short notice. However, you may have to repeat this process whenever the team gets new members or when members need a refresher.
2. Create a standard operating procedure
The other option is to create a standard operating procedure (SOP). An SOP is a document with clear-cut instructions on how to carry out certain processes. It sets the standards for the team members to follow and makes it easier for them consistently perform at a high level.
Compared to conducting training sessions, creating SOPs require a higher level of effort. However, they can serve as a more permanent reference that data scientist/analysts can always peruse when needed.
Tip 6 - Anticipate Questions
When presenting visualizations, analysts need to be prepared to respond to questions that the client may put forward. Clients can bring up anything that piques their interest, including the more minute details.
While it's ok to not have an answer to every question, it's not a good look when you can't comment on the more obvious findings in the visualization. Saying "I don't know" excessively also gives the impression that you haven't sufficiently explored the data.
To avoid this, it's best to spend some time examining the visualization in anticipation of topics that might be brought up. Specifically, analysts should proactively look for the following:
- The main takeaways from the visualization
- Noteworthy trends and patterns
- Outliers
Conclusion
I don't wish to make light of the technical skills needed in data visualization since they are absolutely essential. However, studying BI tools and learning the best practices in design alone won't develop the skills needed to successfully execute a data visualization project.
In reality, this endeavor requires constant communication with clients and team members alike; data scientists/analysts will spend much more time talking and writing than creating actual charts.
The 6 aforementioned tips are my main takeaways from 4 years of creating and presenting visualizations. Following them has enabled me to build effective visualizations with a pain-free experience. Hopefully, they will be of some help to you too.
Thank you for reading!
Subscribe to DDIntel Here.
Visit our website here: https://www.datadriveninvestor.com
Join our network here: https://datadriveninvestor.com/collaborate