Most dashboards stop at the first chart.
They show you a number: points per game, the percentage of visitors wearing a mask, or how many babies got a specific name. While that information is useful, it is where analysis begins, not where it ends.
The application dashboards that truly catch the eye of the Data School selection committee keep digging. They don't just report data; they interpret it.
To get your application noticed, you need to understand the three layers of data depth:
First Order: What Happened?
This looks like reporting raw counts, basic metrics, and simple totals.
Second Order: Compared to what?
Here we start adding context, normalizing data, and calculating percentiles.
Third Order: So What?
At this point we're drawing conclusions, uncovering patterns, and recommending actions.
Businesses don’t hire analysts because they want more charts; they hire analysts to help them make better decisions. Here are three successful Data School applications that show exactly how to move past basic reporting.
Put the Number in Context and Answer "So What?"
A number by itself rarely tells the whole story. Second-order analysis changes the question from "What is this number?" to "How should I interpret it?" This usually means comparing it to a baseline, choosing a better metric, or adding context that alters its meaning.Once you’ve explained what happened and why it matters, there’s one final, crucial step: What should someone actually do with this information?
Example 1: Stosh Sawicz’s NBA Player Comparison
Instead of using raw field goal percentage, Stosh utilized Effective FG%, which properly weights the value of three-pointers.
Instead of simply stating a player averages 30.12 points per game, his dashboard illustrates that this volume puts them in the 99.9th percentile, with zero players ahead of them. Furthermore, by plotting career totals against years of experience rather than calendar years, he allowed players from completely different eras to be compared fairly.
Example 2: Megan Harrington’s Baby Names Dashboard
Megan’s project explored an article from The Atlantic suggesting that modern parents increasingly choose uncommon names to help their children stand out.
Rather than simply plotting the sheer volume of babies given each name, she converted those counts into shares of the total population. She visualized the popularity of top names shrinking over time as rarer names became more common, tracked how often the top five names shifted year-to-year, and mapped the dominant name in every state.
Example 3: Lisa Hitch’s NYC Parks Mask Use Dashboard
Lisa segmented mask-wearing in New York City parks by age, activity, day, hour, and borough. Looking across these dimensions revealed something unexpected: people were actually more likely to wear masks during crowded, social times—like lunchtime and Saturdays. This directly challenged the assumption that people become less cautious when relaxing. She also discovered that seniors were less likely to wear masks while socializing.
Lisa’s dashboard didn't stop at these observations. It asked what they implied for real-world strategy:
- Pivot the Messaging: Instead of funding campaigns focused on why people don’t wear masks, public outreach could reinforce the positive behaviors people already exhibit (e.g., protecting friends during an outdoor lunch).
- Targeted Outreach: Public health initiatives could specifically target social spaces frequented by older adults.
- Smart Distribution: Mask supply logistics could be dynamically adjusted by borough based on observed behavior.
Not every project has to end with a definitive business recommendation. Sometimes the right conclusion is identifying a hidden risk, pointing out a market opportunity, or suggesting exactly what deserves further investigation. The goal is to help your user understand what the findings mean beyond the boundaries of the chart.
One More Thing: Great Analysts Share Their Limitations
One thing all three of these successful applicants did incredibly well was explain what their data could not tell them.
- Stosh noted that historical NBA records before 1976 were excluded due to data availability.
- Megan pointed out that the Social Security Administration suppresses names with fewer than five occurrences, limiting absolute claims about truly rare names.
Good analysts know that uncertainty doesn’t disappear just because you leave it out of your dashboard. Being transparent about your data's limitations builds trust in the conclusions you can support.
Before You Hit "Publish" on Your Application
Before you submit your viz to the Data School, run it through this quick four-question litmus test:
- What happened?
- Compared to what?
- So what?
- What can my data not tell me?
If your dashboard only answers the first question, you’re just reporting. If it builds answers for all four, you’re doing real analysis—and that’s how you land a seat in the next cohort.