DataWonder

Inspiration

Business data is often trapped inside CSV files because analyzing it usually requires spreadsheet expertise or BI tools such as Power BI and Tableau. We wanted to make analytics feel as simple as uploading a file: no predefined schema, no manual field mapping, and no need to be a data analyst.

What it does

DataWonder turns any uploaded CSV into an adaptive business dashboard. It automatically detects numerical, categorical, datetime, and text columns, then generates only the analyses the data supports.

Users can explore an Auto Dashboard, build extra charts in the Chart Generator, select the most useful KPIs and insights, arrange components in a customized dashboard, apply themes, and export the final result.

How we built it

We built DataWonder with Streamlit, Pandas, NumPy, and Plotly.

Pandas and NumPy clean, profile, and analyze uploaded datasets.
Plotly generates interactive, theme-aware charts.
Streamlit powers the upload workflow, dashboard UI, filters, session state, and customization controls.
A shared theme manager keeps in-app charts and exported HTML, PNG, and PDF dashboards visually consistent.
Drag-and-drop layout state stores the user’s selected components, columns, and positions for export.

Challenges we ran into

The hardest challenge was supporting arbitrary CSV files without assuming fields such as Sales, Date, Product, or Customer ID.

We also had to ensure the dashboard never invents insights when a required field is missing. Instead, DataWonder skips unavailable analyses and explains why.

Another major challenge was preserving the exact customized layout in exports. We needed to prevent charts, tables, and long text from overlapping or overflowing into other dashboard columns.

Accomplishments that we're proud of

Built a fully dataset-agnostic analytics workflow.
Created automatic EDA, data-quality reporting, trend detection, correlations, distributions, and anomaly insights.
Added a Chart Generator for users who want more control than the automatic dashboard provides.
Created a visual dashboard customization flow with selectable KPIs, charts, summaries, insights, themes, and drag-and-drop placement.
Ensured exported dashboards preserve the user’s selected content, layout, and theme.

What we learned

We learned that making analytics accessible is not just about generating charts. Users also need confidence that the analysis is grounded in their actual data, clear explanations of what is unavailable, and control over how the final story is presented.

We also learned how important state consistency is: filters, themes, chart edits, selected components, layouts, and exports must all reflect the same dashboard state.