Designing with Dalgo : Enabling AI-powered data insights for non-profits

Dalgo is an open-source data platform aimed at empowering NGOs by simplifying data management. It automates data consolidation, transformation, storage, and visualization into a unified platform, allowing non-profits to fully leverage their data.

Problem Statement

Non-profits collect vast amounts of qualitative data through feedback, surveys, multiple chatbot interactions, & focus group discussions (FGDs).

• Manual analysis of qualitative data is time-consuming and limits insights.
• No direct way to use LLM prompts within Dalgo.
• Users can’t combine qualitative and quantitative data for detailed analysis.

Proposed solution

To solve this, the Dalgo team and the thinkers aimed to integrate LLMs (GPT-4) into DALGO for exploratory data analysis because this would allow users to apply predefined or custom prompts to both quantitative and qualitative data, so the key features mainly include:

1. Custom SQL Filters: Users can specify columns, filter values, or join multiple tables for targeted analysis using SQL commands.
2. Predefined and Custom Prompts: Users can select predefined prompts (e.g., “Sentiment Analysis,” “Areas of Improvement”) for qualitative insights or create custom prompts for more complex analyses.
3. Data Output: The UI directly displays results and offers options to save outputs to the database, download them, or copy them to the clipboard.
4. Data Privacy: A signed declaration from the Account Manager is required to authorize sharing data with OpenAI, ensuring privacy is maintained.

Design process breakdown

Discover

Data collection and research: The design team initiated conversations with the DALGO team to gather insights into their needs and expectations for the LLM feature, and additionally, we also referred to a detailed requirements document provided by DALGO, which outlined all the essential functionalities.

Ideate

Sketches and low-fidelity wireframes: The design process started with rough sketches, leading to low-fidelity wireframes, which we tested and iterated upon based on the feedback.

Design

Final Designs: Following further discussions with the DALGO team, we developed the final high-fidelity designs. In this process, we incorporated user insights and addressed pain points identified in earlier iterations. As a result, these designs strike a balance between usability and functionality, thereby enhancing the overall user experience.

Design Iterations

Initial concepts : Wireframe #1

In the first iteration, we split the screen into a 50-30-20 layout:

• 50% of the screen is dedicated to results/output.
• 30% for filtering and input fields.
• 20% for historical session data.

This design enabled users to concentrate on output, but it also had some drawbacks. For instance, displaying too many input fields made the interface overwhelming. Additionally, collapsing the sidebar for full-screen analysis complicated navigation.

Challenges Identified:

• Too many input fields: The sheer number of filters made data entry difficult.
• Collapsed sidebar: Users struggled to navigate between the LLM module and other platform features when the sidebar was hidden.
• Historical session listing: Though useful, it consumed space that could have been better utilized for analysis.

Design refinement: Wireframe #2

To address these issues, we iterated on the design, opting for a layout that reduced clutter and balanced input and output areas.We split the configuration panel and data output into two connected sections, with both areas given equal weightage.

Issues Identified:

The input area appeared too cluttered, with an overwhelming number of fields, making the user experience less smooth.

The interface did not emphasize crucial elements like the prompt selection, which is vital for user interaction.

Finalized version, optimized for better UX: Wireframe #3

In the final iteration, we significantly reduced the number of input fields while maintaining critical functionality, and the main priorities were readability and ease of use, so the key changes include:

1. Larger Output Area: The output section now occupies most of the screen, providing more room for displaying detailed insights from LLM analysis.
2. Simplified input fields: We responded to client feedback by minimizing the number of input fields, ensuring that users can filter by SQL commands without feeling overwhelmed by options.
3. Visible Sidebar: The sidebar remains visible throughout, allowing users to easily navigate between different sections of DALGO without having to exit the LLM analysis page.
4. Historical Sessions: Users can access saved analysis sessions with one click, improving workflow efficiency.

Solutions Implemented:

The final version achieves a balance between usability and functionality. We reduced the input fields and increased the focus on the output, resulting in a more intuitive and user-friendly design without compromising the platform’s powerful data analysis capabilities.

We improved the navigation by keeping the sidebar visible, which made the entire experience more unified and efficient.

Conclusion

The LLM Data Analysis feature on DALGO went through multiple design iterations, each improving the user experience while staying true to the platform’s goal of providing non-profits with powerful, AI-driven insights. Additionally, by refining the layout, minimizing unnecessary inputs, and focusing on the user journey, we created an intuitive and effective solution for nonprofits to harness their data for impact analysis.

Watch out this space for more interesting blogs and content from us. Stay tuned…