Perceived Performance

Eighty Days App UX Case Study: Instant AI Feedback with Response Buffering

Implementing a buffering technique to deliver AI results instantly, reducing user wait time and improving satisfaction.

A mobile screen showing text instantly streaming into a chat window.

Impact by the Numbers

By prioritizing Time to First Interactive (TTFI), we transformed a slow waiting experience into an engaging, dynamic conversation.

Time to First Interactive (TTFI)

90% Reduction

Perceived Wait Time

Negligible

User Frustration Score

Reduced

Overall User Satisfaction

Highest

The Challenge

The AI Latency Problem

AI Engine Requirement

The 'Eighty Days' app relies on complex AI generation (OpenAI APIs) that includes 'thinking' or processing time to generate detailed, structured outputs (like JSON itineraries).

The Core Problem

Traditional API calls require waiting for the entire response to be generated and validated before displaying anything to the user, resulting in a frustrating and lengthy silent wait.

The Roadblock: The Silent Wait

Slow Overall Execution Time

Due to the complex 'thinking functionality' required, the total time for the AI to finish and send the complete response was long.

Total Response Time

5 - 10s

Zero Feedback User Experience

For the entire duration of the AI processing, the user saw no output, creating anxiety and the perception that the app was frozen or slow.

Time to First Byte (TTFB)

High

Data Fragmentation Risk

Receiving partial, buffered data could lead to invalid or broken JSON objects if not handled meticulously on the frontend.

Data Integrity Risk

High

The Client’s Vision: Instant, Dynamic Feedback

A user watching a response stream into a chat window.

Minimize Perceived Latency

Reduce the time the user spends looking at an empty screen, even if the total processing time remains constant.

Utilize Buffered API Data

Implement a method to stream data chunks from the OpenAI API as they are generated, rather than waiting for the final package.

Maintain Data Integrity

Develop a robust frontend mechanism to correctly piece together partial, buffered JSON objects into a valid final output.

Our Solutions

Instant Gratification via Streaming UX

Reduced Time to First Interactive

By immediately displaying the first text buffer, we drastically lowered the TTFI, making the app feel responsive and fast, even though the total server time remained the same.

Frontend JSON Reconstruction

Our innovative client-side mechanism ensures that every partial, fragmented data chunk is correctly parsed, validated, and merged into a complete JSON itinerary without breaking the application.

Enhanced Perceived Performance

The flowing, progressive text output turns a period of waiting into a period of engagement, managing user expectations and significantly boosting the perceived speed.

Optimized User Experience (UX)

This solution prioritizes the user's perception over raw execution metrics, delivering a modern, high-quality experience expected from state-of-the-art AI applications.

Project Visuals

Behind the Transformation

Key Takeaways

Main Learnings & Highlights

Perceived Speed Trumps Actual Speed

The project validated the principle that user satisfaction is driven by perceived performance. Giving immediate feedback, even if partial, is more valuable than a faster total response time that involves a prolonged silent wait.

Frontend Processing is Crucial for AI UX

Handling complex data processing, like reconstructing fragmented JSON objects, on the client side is a necessary technical investment to deliver a smooth, streaming user experience for AI outputs.

UX is About Managing Expectations

The buffering technique is a prime example of using UX to manage user expectations. By showing the output is in progress, the system avoids the frustration associated with ambiguous loading states.

Limitations That Guided The Process

Project Timeline

The project carried a firm deadline, pushing the team to stay aligned and deliver on time.

Time taken

125 Hrs

Phase 1

Initial planning and research took place, establishing the framework for the project.

Time taken

50 Hrs

Phase 2

Development and implementation, where the core features were built and tested.

Time taken

75 Hrs

Phase 3

Final review and adjustments, ensuring everything met the quality standards before the launch.

Time taken

30 Hrs