Data Visualization

Enhance the search and preview functionality across multiple files.

Role

UX Designer (Design challenge)

Team

‍Myself

Duration

3 days

Design problem

Kaggle's current preview feature falls short when users need to search across multiple files.

Let's look at this YouTube video statistics dataset.‍

‍Using the current preview experience, here are data questions that cannot be easily answered like Is there a video that appears in both the US & CA? or How many unique videos are there in Europe in total?

Research

What I found

Compared to experienced data analysts, novices face more challenges.

Research shows that experienced data scientists did not get border for the data preview they do not mind since they have access to advanced tools. However, novice data analysts who lack knowledge or experience in using advanced analysis tools usually struggle the most.

HMV improve search and preview capabilities across multiple files to assist novice data analysts, helping them quickly find the information they need?

User journey map

Current users pain points and opportunity.

After gaining a better understanding of the design goal, I developed a user journey to identify the specific challenges novice data analysts face in the dataset preview.

Final solution

Data Preview on data visualization Analysis

Users can leverage data preview features to quickly assess datasets, leading to more informed data visualization analysis.

User journey map

New user journey

After testing with potential users, here is the new journey map that users will use with the updated data preview function.

The process

Competitor analysis

This is a time constraints design challenge, my design approach involves taking a closer look at how competitors operate to draw inspiration.
  • Do they provide dataset or file previews on their site?
  • How do they display and organize multiple files in the dataset or file preview?
  • Is their approach to multiple file previews user-friendly and conducive to exploring data and discovering insights?
Insights
Competitor 1
Competitor 2

1. In-build data analysis preview tool.

  • Easy to use, neat design.
  • Showing graphics.
  • Allow user to optimize.

2. Users need base DA knowledge and skills.

3. Competitors didn't provide direct data/file summary apart from the data graphics.

All dataset files are visually preview at once.

‍kaggle: Only one file is expand; click "Data Explorer" to see others files.

  1. Quicker way to scan all the data.
  2. Easier to catch the insight than only one file.

Still can’t find an insights directly if it need two files to analysis.

In-build data analysis preview.

  1. Quickly get insights thought simple data analysis tool.
  2. Easy to use & neat.

Still require base knowledge of data analysis since the analysis configuration need manual input.

File is preview in content. Given necessary information, but set a high bar for novice DA users.

In-build auto-generated data analysis.

  1. Effortless get the insights.
  2. Less technical knowledge require.

But...

  1. Some dataset still need optimize to get the answer.
  2. No summary, only graphs.

C

Design exploration

I then crafted potential solutions based on the earlier insights.
a.
Add auto-generated interactive visualization tools on data preview (Graphics + Summary).
✅Easy & efficient.

✅Allow users to get insights either from graphic or context summary.
⚠️Users might select the unrelated file, resulting in an unrelated answer - increasing cognitive load.

b.
Utilize AI chat for more goal-directed.
✅Most goal-directed.
✅Easy & efficient.
⚠️High cost.
⚠️When datasets contain many files, automatic identification and analysis can lead to prolonged processing times.

C

Design decision

Considering Kaggle's prioritizes user engagement and ease of use, and users are more comfortable with visualizations (85% think data preview is useful), solution a. would be a better fit.

Since user engagement and ease of use are top priorities, solution a. with auto-generated interactive visualization tools seems more suitable.

However, it's essential to address the risk of users selecting unrelated files. Implementing clear instructions or file selection safeguards could mitigate this issue.

C

Iteration

The solution to address the risk of users selecting unrelated files can involve adding search keywords for datasets and providing file previews during analysis.
1. Add searching key worlds for dataset files.

Input the keywords on the search box.

System auto-highlights the files which involve keywords.

2. Files previews during analysis.

Click the file.

Show the file details for preview.

What I learned

Epilogue and reflection

Key learning:

1. Quickly identifying design solutions is essential for completing projects independently.
2. Recognized that seeking help is equally important when working alone.
3. Limited research scope and time constraints impact comprehensiveness.

© 2025 Junwen Yu. All rights reserved