Data practitioners know that a critical first step in data analysis is taking a careful look at their data in exploratory charts and tables. They also realize the value of building clear, compelling visuals for end products like dashboards and reports. So when it comes to keeping a close eye on their data, analysts have the bookends covered.

But something unfortunate often happens in the in-between, when analysts are heads-down in data cleaning, wrangling, modeling, troubleshooting, and iterating: data visualization falls by the wayside. 

An illustration shows charts for exploratory data visualization on the left, and a final dashboard on the right. Between is a blank area labeled "data visualization desert" with a tumbleweed, representing the lack of data viz in the middle of analysis.

Visual data analysis merges data visualization and analysis, helping viewers to explore and answer questions about complex data through charts, tables, and graphics. That’s valuable throughout the entire analysis process, not just at the endpoints. 

Here, we share how end-to-end visual data analysis creates more opportunities for discovery, supports cross-functional collaboration, and reveals mistakes before too much damage is done. 

Create opportunities for on-the-fly discovery

In the messy middle of data analysis, new patterns and anomalies can be revealed as you wrangle, aggregate, and analyze data in different ways. This can inspire questions and exploration paths that you hadn’t previously considered. If you’re not visualizing the data at each step, however, those discoveries can remain hidden in overwhelming tables, or worse — entirely unviewable, only existing ephemerally within an intermediate line of code. 

Manually building visualizations to explore the output of each data processing step can be prohibitively time consuming. Hence, the data visualization desert. 

A good alternative is to use tools that provide visual summaries of the data at each step in your analysis by default. For example, in Observable Canvases, all data processing (whether performed in SQL, or using UI options for common operations like joins and aggregation) are accompanied by a summary table containing small charts in each column header that provide a quick view of variable distributions. 

A summary table where a concise data visualization is shown atop each column to give a quick view of that field's contents and distribution.

Whether you use code or UI options in Observable Canvases, a summary table with concise data visualizations makes it easy to explore the shape and content of the data at that stage in your analysis.

From there, it just takes a few clicks to make larger, multivariate charts that let you dig deeper into the patterns you’ve discovered, which could yield unexpected and valuable insights for your business.

Improve accessibility and interpretability

The data analysis process is often inaccessible to cross-functional collaborators like business and product managers who may lack familiarity with the programming languages or analysis tools used by your data team. And working openly, for example by sharing your code on GitHub, is a far cry from really empowering them to participate in your work. 

Nowadays, most people have at least a basic level of visual data literacy. They’re constantly exposed to data visualizations in the news, through fitness tracking apps, while digging into unexpectedly high electricity bills, and more. Visualization has become somewhat of a universal language that allows many people — regardless of technical knowledge or skills — to make sense of data. 

So, if you want to help your non-coding collaborators to review and interpret analyses and give high quality feedback, visual analytics is a great way to go. 

When collaborators visually follow how data is transformed throughout analysis, they can confidently interpret what they’re viewing at any point. For example, they could see how data has been filtered and aggregated upstream, to know exactly what data is included in a larger downstream chart. 

Tools that visualize data transformations and operations throughout analysis, as shown above in an Observable Canvas, let cross-functional collaborators (including those without coding experience) confidently review and interpret results.

Notice mistakes and anomalies early on

In business analytics, mistakes are both inevitable and potentially costly. 

Unit conversions get fumbled. Whole groups inadvertently disappear with the smallest typo (was it == “blueWhale”, or == “bluewhale”?). Oh, and that zip code? Yeah…at some point that got converted to a number, when it should have been treated as a nominal variable in your model. 

The types of mistakes described above won’t show up as error messages in the console, since code doesn’t know (or care) if what you’re asking it to do with the data is actually correct. If you’re lucky, they’ll be caught in code review by an eagle-eyed colleague, or while diagnosing obviously wrong outputs. Unfortunately, data processing mistakes don’t always reveal themselves so easily. 

Visualizing data throughout analysis is an additional way that you and your collaborators can catch mistakes. With more eyes on the data the whole way through, you’re more likely to identify issues with how it’s processed. And, since your team can visually track changes to the data at each step, it’s easier to determine exactly where in the analysis things went wrong. 

A high resolution density chart reveals a large gap in wind speed records. A manual annotation points out the gap, with text saying "I don't think we accounted for this gap."

Visualizing data throughout your analysis increases the likelihood of finding mistakes and noticing anomalies (like missing data, as shown above) earlier in the process.

Finding mistakes may sound like a bad thing. But, if it happens early enough in the process, it’s a gift. Visual data analysis can increase the likelihood of catching mistakes and revealing red flags, before it’s too late. 

Don’t let data out of your sight

End-to-end visual data analysis helps you to keep eyes on the data throughout analysis, ensuring that surprises — whether exciting, or concerning  — don’t slip by unnoticed. 

By using tools that make it quick and easy to see your data throughout the analysis process, instead of just at the bookends, you can maximize visibility without slowing down for more accurate and interpretable insights. 

Learn more about fast, visual data analysis in canvases, and sign up for early access today.