AI diagnostic Module
building trust for Meituan Merchants
Built from 0 to 1 + Iteration: a high-trust, AI diagnostic framework to bridge the professional advertising gap. Achieved a 23% increase in adoption rate and a 74% reduction in operational time for Small Business Owners.
As a Product Design Intern on Meituan’s Commercial Value team, I was responsible for the design of new features and iterations for the AI-powered advertising diagnostics module on the merchant advertising platform.
The project focused on using AI and multi-dimensional data to generate personalized advertising recommendations for merchants. I drove the end-to-end design from 0 to 1 MVP launch through subsequent iterations, covering the full funnel including entry exposure, AI insights presentation, conversational interaction, and decision-making for recommendation adoption.
The MVP and subsequent iterations were successfully launched. The MVP reached 23,000+ clicks in its first month with a limited merchant rollout, and the second iteration target to improve recommendation adoption by 20 - 30%, contribute to higher advertising conversion.
May 2025 - Aug 2025
Internship Project
Skills:
Product Thinking
UX Prototyping
Visual Design
AI Design
Iterative Optimization
Tools:
MasterGo
Timeline
01. Problem Define
Contextualizing merchant friction and business growth challenges.
Contextual Challenges
Merchant Friction
66% of Merchants find manual ad settings "too complicated" due to high industry barriers and limited literacy.
Market Pressure
Competitors achieved a 12% conversion lift via AI, shifting the industry benchmark and necessitating defensive innovation.
Business Growth
Platform revenue growth decelerated as low campaign ROI led to a decline in merchant confidence and high operational support costs.
Product Strategic Challenge
How Might We leverage AI diagnostics to bridge the advertising expertise gap for small business owners, empowering them to optimize ad performance while driving platform revenue growth and reducing operational costs?
02. Competitors Research
Synthesizing industry standards to define AI interaction patterns.
Strategic Utility
IA and Visual Identity
Develop Co-pilot persona to balance authority with empathy.
Utilize Semantic highlighting and modular cards to ensure key metrics are scannable in seconds.
Implement Progressive Disclosure to show logic while prioritizing outcome-oriented data.
Interaction Feedback
Integrate Micro-feedback loops (Sentiment icons) to reinforce the Human-AI collaborative bond.
Control the information flow speed during generation to prevent user disorientation from sudden data dumps.
03. MVP Design Outcome
Launching the core diagnostic flow with frictionless execution
Main User Flow
Users enter the AI Diagnostic tool from the homepage, where the system progressively generates specific ad suggestions. Merchants can then choose to apply all optimizations with one click or manually adjust individual settings.
Entry Point
High-Saliency Awareness
Visual Saliency ensures discoverability
Information Externalization triggers curiosity and highlights immediate value
Brand Identity through AI persona
Progressive Loading
Cognitive Pacing
Managing Perceived Time
Algorithmic Transparency increases user trust
Diagnostic Result
Strategic Persuasion
Semantic Scannability highlight key metrics
Data-Driven Confidence directly present tangible business gain
Low-Friction Interaction with preset questions and "One-Click Auto-Optimize"
Contextual Adjustment
Zero-Friction Modification
Preserving Context by eliminating page-jumping when making adjustments.
Reuse familiar setting components reduces the learning curve
Post-Action Feedback
Value Validation
Instant visual confirmation provides psychological closure
Post-Action Feedback
Value Validation
Instant visual confirmation provides psychological closure
Status Variance （Low-fi)
From MVP to Strategic Refinement: Optimizing for Impact
Moving from a basic diagnostic tool to a persuasive AI Co-pilot. By aligning design details with business growth, we re-architected the V2 experience to drive deeper trust and effortless execution.
04. MVP Assessment & Opportunity Mapping
Identifying the core bottleneck through real-world adoption data.
-
45%
Deviation from Target Adoption
Initial real-world adoption fell significantly short of our projected growth baseline, signaling deep-seated execution friction.
<
3%
Full-Strategy Adoption Rate
While merchants engaged with the tool, less than 5% completed the full optimization path, missing out on the cumulative ROI.
70%
-
Single-Item Selection
The vast majority of active users treated suggestions as isolated tasks rather than a cohesive business strategy.
The "Selection Bias" Paradox:
High entry traffic was met with low execution depth. Our audit revealed that merchants were "Browsing but not Buying" into the full AI strategy.
The New Mission
05. Understand User
Second-phase user interviews to decode the psychological barriers to adoption.
Interviews
"Sometimes AI suggestions are designed to benefit the platform rather than my shop. I don't know what can I return, it feels like a trick to make me spend more."
— Trust Barrier
"I’m a chef, not a marketer. I stick to the most 'conservative' settings because I don't know how to fix a failing campaign. I just want more customers"
— Expertise Gap
"Between the kitchen and taking orders, I only have a 5-minute window to check my phone. If it isn't instant and effortless, I’ll just skip it."
— Efficiency Crunch
"Setting up number for bidding is always hard as I don't know what others in my neighborhood are doing, so how can I stay competitive?"
— Need for Benchmarking
What I learned:
Cognitive & Operational Accessibility
Bridge the literacy gap by Intuitive Modular Information Hierarchies.
Utilize One-Page UI and Single-Click Adoption flow to minimize cognitive load and ensures immediate execution.
Transparency & Trust Alignment
Eliminate the "Black Box" effect by visualizing data sources and Peer Benchmarking.
Prioritize Logic Traceability to establish AI credibility.
Value Perception & Sustained Retention
Utilize Value-First Architecture to catalyze initial adoption through tangible ROI projections.
Balance systemic efficiency with User Agency to provide flexible, human-in-the-loop optimization framework.
Design Strategic Challenge
How Might We transform opaque AI ad diagnostics into transparent and high-trust decision framework that empower merchants to execute with low-friction and strategic confidence?
06. Design Stretagy for Impact
Solving for conversion by addressing the four dimensions of the "Execution Barrier."
(01)
From Task Lists to Holistic Solutions
Problem: Fragmentation
Solution: Holistic Strategy Packaging
The suggestions list were presented as isolated bubbles. This fragmented layout conveys them as independent, optional tasks rather than a holistic strategy.
Grouped all suggestions into a unified solution with a single "Adopt All" primary action.
(02)
From Information Dump to Value Perception
Problem: Value Obscurity
Solution: Directive Copywriting & ROI Projections
Merchants hesitated because they couldn't see the "Bottom Line." Without clear revenue expectations, they lacked the incentive to change their settings.
Replaced vague titles with action-first directives and highlighted predictive revenue lift prominently.
(03)
From Black Box to Strategic Trust
Problem: Opaque Logic
Solution: Logical Traceability & Data Evidence
Merchants were skeptical of the "Black Box" AI suggestions. A lack of reasoning led to a Trust Gap.
Exposed the reason behind. We integrated specific data evidence to validate each suggestion.
(04)
From Cold Tool to Empathetic Co-pilot
Problem: Low Affinity
Solution: Dynamic Persona & Situational Empathy
The static, head-only avatar felt like a rigid system bot, failed to resonate with the money-making goal of merchants.
Upgraded to a full-body animated character with business-aligned motions like "Coin-Toss" to enhance brand affinity.
(05)
From Static Reading to Hands-Free Accessibility
Problem: Operational Friction
Solution: Multi-modal Interaction (Voice-over)
Busy merchants in fast-paced kitchen environments often find it impossible to read long text-based advice on a screen.
Integrated Voice Broadcast (TTS) to narrate the diagnosis and optimized the visual hierarchy for quick scannability.
07. Iteration 2 - Design Outcome
Measuring success through post-launch data and merchant sentiment analysis.
07. Validation & Business Impact
Measuring success through post-launch data and merchant sentiment analysis.
As the feature launched shortly after my internship concluded, post-launch metrics were not fully available. The design was therefore evaluated against predefined success metrics and target improvements.
Success Metrics
• Recommendation adoption rate
• Interaction rate with AI insights
• Completion of optimization actions
• Merchant trust and feedback scores
Target Outcomes
• +20–30% increase in recommendation adoption
• Higher engagement with AI insights
• Improved merchant trust in AI recommendations
• Clearer path from insight → action