Building Data Applications the Modern Way

Data powers nearly every decision in modern organizations. But how we interact with that data, how we make it useful, has changed dramatically. At the Data Driven Wisconsin 2025 conference, I shared some of the lessons I’ve learned building tools and platforms that help teams create data applications that make an impact where it counts.

Data applications, in our view, are the most practical way to connect data teams with decision makers, because they open up a wide range of possibilities for making analysis useful in practice.

What are data applications?

At their core, data applications are pieces of software that guide business decisions using data. They nearly always include visualizations and are often tailored to a specific business problem. Data apps play an important role in translating data into actions.

You probably interact with many data apps every day:

Dashboards are data apps for tracking metrics over time
Reports are data apps analyzing a snapshot of data with low interactivity
Many spreadsheets are data apps prioritizing flexibility rather than presentation

The most powerful apps go beyond these general situations by enabling exploration, prediction, and even action. Building data apps starts with your choice of the tools available to you.

The tools we reach for

When you open a CSV file or connect to a dataset, your tool choice determines the analysis you can perform and the insights you can extract. Each option trades off between speed, flexibility, and technical requirements, from built-in dashboards that work immediately to custom applications that solve specific problems.

Built-in dashboards: Many tools include basic analytics right in the interface
Spreadsheets: Quick, accessible, and perfect when all you need is a simple chart
Point-and-click SQL tools: Add filtering, access to databases, and sharing across teams
Python notebooks: Unlock custom processing, machine learning, and complex visualizations
“Vibe coding” tools: A newer category, blending the power of code with the usability of interactive interfaces
Coding applications: Building entire applications to understand a specific set of data

These tools exist on a spectrum, and choosing between them requires understanding what you gain and lose with each option.

The tradeoffs we make

When choosing a tool or platform for data apps, the tradeoffs tend to fall into three big categories:

Features: customization, interactivity, and scalability.
Costs: learning curve, development time, and long-term maintenance.

Most users end up selecting the tool they are most comfortable with. However, a more general approach is to balance the tradeoffs. I always recommend selecting the simplest tool that gets the job done. The main dimensions to consider are:

How big is the data?
Spreadsheets generally get harder to use above 10,000 rows.
How sophisticated are the outputs?
Point-and-Click tools often have many output options, but not as many as custom code would.
Who is the audience?
Python notebooks often aren’t C-Suite ready.
How much will it be used?
High volume reports can justify investing developer time to make sure they’re perfect.

Another approach is to try every tool in increasing complexity. Since more sophisticated tools take significantly longer to set up, the time spent in a less sophisticated tool likely isn’t that high. For example, exploring data in your CRM system might follow something like this:

(15 minutes) Look at the data in the CRM first.
(30 minutes) Export the data to CSV and open it in Excel for exploration.
(4 hours) Load data in a database to create a C-Suite ready dashboard in PowerBI.
(2 days) Take the data into a notebook to answer a complex follow-up question.
(2 weeks) Code a custom tool to present the deeper findings.

When doing this, it’s often useful to keep in mind that AI tools can significantly accelerate the work in some of these tools and hence justify skipping some less sophisticated tools. For example, in Plotly Studio you can get a lot of the deeper analysis a custom notebook offers in the time Excel exploration typically takes. Ultimately the deciding factor is the quality and features of the data app you need to produce.

Why the modern way matters

The best modern data apps are the right ones for the job. Sometimes you need a quick answer from reviewing the data in a spreadsheet. Other times you need a sophisticated application to analyze data as it streams in. Ultimately though, every data app needs to strike a balance: powerful enough to answer your question, but approachable enough for the user to understand.

Choosing the right tool for each job is critical. With everyone trying to do more, faster, and with fewer resources we need the ability to create better data apps with less effort. This also means more business experts just want to understand their data directly rather than waiting for an analysis or a report from another team.

This is why we’ve seen the rise of platforms like Plotly Studio, which aim to give teams the ability to combine the best of both worlds, flexibility and accessibility, without locking them into rigid tools.

What lies in the future of data analytics apps

As the data and AI evolves, I see three trends shaping the future of data apps:

Self service analytics - the best tools will empower teams of all skillsets and expertise to build and share insights.
Seamless integration - data apps will connect to all your data with minimal effort using AI generated connectors.
Predictive analytics - data apps will guide decisions.

The future will very much be shaped by the flexibility that LLM development enables.

Self-service analytics as a whole has historically struggled from balancing the sophistication of the possible analysis and technical ability. This is no longer an issue when AI can provide much of the necessary expertise. The experience is closer to requesting a report from an analyst than self-service analytics have ever been in the past. This external expertise also unlocks advanced analytics and predictive technologies that were previously reserved for only the most critical problems, such as bringing machine learning expertise to every problem.

Similarly, access to data has often thwarted efforts to empower business experts to analyze their own data. Often the data needed to answer a business question is buried in a highly technical database, and hence a data engineer is needed to retrieve it. With LLMs, the technical barriers to data access are much lower since they can search for and retrieve the data. This makes data governance more important than ever, but ultimately enables non-technical users to do things they never could before.

Building data apps the modern way is a bit more than choosing the “coolest” new platform. It is an opportunity to understand your skills, weigh the tradeoffs, and build applications that actually help your organization make better decisions.

That’s the opportunity in front of us: to move beyond static reports and into a world where data apps are dynamic, interactive, and integral to how organizations run.

So the question is: how is your organization rethinking its approach to data apps?