top of page

Meal Map

Role: 

User Researcher, UX Designer

Meal Map is an accessibility-focused AI grocery assistant designed to simplify meal planning and grocery shopping. The prototype uses AI to generate personalized meal plans based on dietary needs, budget constraints, and user preferences, then automatically builds a grocery list and guides users through the store using optimized shopping routes and aisle-level navigation. 

The goal of the project was to explore how human-centered AI can reduce decision fatigue and improve the grocery shopping experience while maintaining transparency, accessibility, and user control. Through user interviews, task analysis, and usability testing, we iteratively refined the prototype to improve usability, trust in AI recommendations, and overall shopping efficiency.

Meal Map Logo.png
518 User Journey Map (1).png

This user journey map illustrates the typical grocery shopping process and highlights the challenges users experience at each stage, from planning meals to post-shopping tasks. The map identifies key user actions, thoughts, emotional responses, and pain points such as decision fatigue when planning meals, uncertainty when choosing stores, difficulty navigating crowded aisles, frustration with coupons and checkout systems, and issues with substitutions or missing items. By visualizing these experiences, the journey map helped our team identify opportunities where AI-powered support could improve the process, such as automated meal planning, personalized grocery lists, store comparison tools, and in-store navigation assistance. These insights directly informed the design of MealMap, guiding the development of features that reduce cognitive workload, improve transparency, and create a more efficient and accessible grocery shopping experience.

This task analysis breaks down the grocery shopping process into six key stages:

​

  • Planning Meals

  • Building Shopping Lists

  • Comparing Stores & Prices

  • Planning & Transportation

  • Navigating & Parking around the Store

  • Reviewing & Adjusting Purchases Afterwards

 

Each stage outlines the specific actions and decisions users typically make, such as checking pantry items, comparing prices across stores, locating products in aisles, and adjusting future plans based on what was purchased or unavailable. By mapping these detailed steps, the analysis helped our team better understand the complex decision-making and cognitive effort involved in everyday grocery shopping. These insights directly informed the design of MealMap, allowing us to identify where AI could provide the most value—such as automatically generating grocery lists, suggesting substitutions, comparing store prices, and guiding users through stores with optimized navigation.

image.png
image.png

This chart shows the frequency of themes mentioned by participants during user interviews, highlighting the most common experiences and challenges in the grocery shopping process. Key themes included meal planning habits, accessibility needs, use of digital tools, price sensitivity, and issues with missing items or substitutions. By identifying which topics appeared most often, our team was able to prioritize the most important user needs. These insights directly informed the design of MealMap, guiding features such as AI-assisted meal planning, accessible interface options, and tools to support price-aware and efficient grocery shopping.

image.png

This heatmap visualizes the sentiment analysis of themes discussed by participants, showing whether experiences related to each theme were generally positive, neutral, or negative. Themes like meal planning and convenience-based shopping were mostly neutral or positive, while issues such as crowds, missing items, and accessibility barriers generated more negative sentiment. By highlighting the emotional impact of these experiences, the analysis helped our team identify where improvements were most needed. These insights informed the design of MealMap, guiding features aimed at reducing frustration, improving accessibility, and creating a smoother grocery shopping experience.

Wireframes

Preference Setup

image.png
image.png
image.png
image.png

In-Store Navigation

image.png
image.png
image.png

User Testing

image.png

This chart compares the number and severity of usability issues identified during testing of Prototype 1 and Prototype 2. After incorporating feedback from the first round of testing, the total number of issues decreased from 72 to 53, representing a 26% reduction in overall issues. Improvements were seen across all severity levels, including a 17% decrease in major issues, a 24% decrease in moderate issues, and a 44% decrease in minor issues, while no critical issues were identified in either iteration. These results indicate that design revisions successfully addressed many usability concerns identified in the first prototype.

image.png

This heatmap shows the distribution of positive, neutral, and negative sentiment before and after the prototype redesign. Before iteration, negative sentiment dominated (70%), suggesting widespread frustration with the initial prototype. After improvements were implemented, positive sentiment increased to 60% while negative sentiment dropped to 10%, demonstrating a substantial shift in user perception and overall satisfaction with the system.

image.png

This chart illustrates the average sentiment expressed by participants before and after the prototype iteration. Initial testing of the first prototype generated an overall negative sentiment score, reflecting frustration and usability concerns. After implementing improvements based on participant feedback, the second prototype produced a positive average sentiment, indicating that users responded more favorably to the updated design.

image.png

This dashboard summarizes several usability metrics collected during testing of both prototypes, including Single Ease Question (SEQ), System Usability Scale (SUS), DRT score, and Net Promoter Score (NPS). The second iteration showed clear improvements, with SUS increasing from 65.8 to 79.5, indicating stronger perceived usability. Task ease scores also improved for both meal planning and in-store navigation, and the Net Promoter Score increased dramatically from –50 to +40, suggesting that users were significantly more likely to recommend the updated prototype.

Challenges

  • Supporting diverse user needs 

  • Maintaining predictability

  • Explaining AI behavior

  • Breakdown recovery wasn’t fully supported

Successes

  • Meaningful usability gains after iteration—SUS improved from ~66 → ~80 (“good” usability).

  • High relevance & fit—meal planning averaged 4.6–4.8 in “fits my process/simplified my shopping.

  • 3D navigation feature positively received

  • NPS improved dramatically—from –50 to +40 (+90 points)—showing a major shift toward user satisfaction and likelihood to recommend the system.

Limitations

  • Limited interaction fidelity made list–map transitions, budget sliders, and dense screens harder to interpret

  • Future Direction - Develop a high-fidelity, fully interactive prototype with smoother transitions and clearer item-to-map binding for better SEQ outcomes

  • Relevance learning is limited and doesn’t adapt to household patterns

  • Future Direction - Improve personalization by integrating budget trends and dietary habits.

Recommendations

UX

  • Clarify substitutions by surfacing accepted replacements directly in the list and map.

  • Simplify dense screens (budget sliders, filters, accessibility options)

  • Support varied shopper roles (novices, neurodivergent users, mobility-restricted users) through adaptive help

AI

  • Improve substitution intelligence with flavor similarity, dietary context, and personal preference learning.

  • Prioritize adaptive routing

  • Instrument and audit AI decisions to monitor friction points, fairness, and model drift over time.

Teaming/Training

  • Maintain predictability through consistent patterns that help users build accurate mental models of the AI

  • Evaluate collaboration quality using trust, workload, predictability, and error recovery—not just usability

  • Support adaptive assistance that increases or decreases guidance based on user experience, cognitive load, or past interaction patterns

bottom of page