Beli Concept: Taste Compatibility Designed by Food Blend

“Where do you want to eat?”

A question that is asked frequently, but typically left at “I don’t know.” You open Beli to look for recommendations on places to eat with your friends. Even after scrolling and searching through the feed, you fail to find a place that you suits both you and your friends’ preferences.

Beli is a recently developed app that allows users to share restaurants, cafes, and desert places with friends. It is a social app made for food lovers.

This leads us to the question: How do we decide where to eat with friends?

What is currently wrong with Beli?

While Beli is successful in providing trustworthy, detailed reviews of restaurants that friends have been to, it fails to guide users to a decision. I wanted to further explore the gap between recommendations and making choices on where to eat — decision fatigue.

Conducting User Research 🔍

When conducting research, I wanted to explore how users use Beli to help them decide a place to eat and discover which social aspect they prioritized.

After receiving a notification from my friend, I opened the app to browse and rank the restaurant we went to together.

When I want to eat with my friends, I use beli and surf the app through the feed to look for recommendations.

Beli encourages me to try new places to eat through the school leaderboard and ranking system.

After interviewing active users of Beli, the main takeaways of my findings were:

1. Friend’s activity and engagement improves and encourages others to contribute to the social side of Beli.
2. Encourages travel and food exploration for people who might live in populated areas
3. Confirmation of trusted reviews of places to help decide where to eat (typically for parties more than one)

Revisiting the Problem 🤔

Based on the insights from my user interview and my original product thinking, I wanted to proceed and focus on the people problem:

When I am deciding on a place to eat, I want to choose a restaurant based on personalized recommendations, so I can confidently choose a place that I will enjoy. But this is hard because:

1. The recommending system is hidden and does not provide sufficient information or context about the recommended places
2. It is hard to connect personalized recommendation system with friend’s activity to make a decision
3. Existing solutions has a place for multiple user input without a specific direction in where to take them

With this in mind, I propose the question:

How do we maintain and prioritize trust while streamlining the process of transforming personalized recommendation to meaningful decisions?

Brainstorming for Solutions 🧠

Adelaide (left) and Claudia (right) collaborating for a brainstorming session.

Before brainstorming, I recruited two of my friends Adelaide and Claudia to come up with ideas. When thinking of ways to address the problem, I identified three opportunity areas to brainstorm more specific solutions:

• Friend’s Data
• Decision Making
• Feed Navigation and Recommendations

From these three opportunity areas, we came up with three solution spaces with two specific solutions in each of the spaces — six solutions total.

Integrating and Implementing Friend’s Data

Since socializing is a huge part of Beli’s foundation, we wanted to explore ways we could utilize the data shared with friends to see how it could drive and encourage engagement.

• A feature that pulls from what friend’s bookmarks and the user’s bookmarks to remix the two to create a list of recommendation
• Food blend (similar to Spotify Blend) where the user can invite friends to explore taste crossovers and a “compete” feature where users can compare the restaurants individually to decide a place to eat

Decision Making

This involves streamlining and simplifying the decisions making process, more specifically targeting decision fatigue. Even with the abundance of choices, it is hard for users to decide a specific place in which they want to eat.

• Q&A process to break down the decision making process into smaller steps — organizing and structuring the experience.
• Specific user tags assigned to people to give themselves a theme or category based on places they’ve visited and things they like

Feed Navigation and Recommendations

This solution space specifically explores how to distinguish the recommendation system to encourage more user interaction. By making it more user-interactive, it can be used to promote personalized food discovery and minimize decision fatigue.

• Separate inbox for receiving recommendations
• Additional filter to the current recommendation system

Analyzing the Possible Solutions 🤹

For each of the solutions, a SWOT analysis & feasibility and impact analysis to contemplate the possible implementation of each solution.

SWOT (Strength, Weakness, Opportunities, and Threat) Analysis was conducted for reach of the solution spaces we brainstormed

Analyzing the feasibility and impact of each of the features

My initial problem was to address the gap between reviews and their decision on where to eat. It was altered to be more specific after Beli developed a new feature that allowed users to request recommendations from their friends.

My new people problem focused on improving the current recommending system by focusing on how group discovery to reduce the decision fatigue that comes with receiving suggestions.

With these points in mind and from the analysis, I decided to focus on was Food Blend** and Plate Off** where users can compare the recommendations that are compatible to come to a decision, addressing the people problem.

Designing the Solution 🎨

Exploring Low Fidelities

Food Blend and Plate Off introduces a new way that users can make lists and compare the restaurants they want to go to. For Food Blend, I wanted users to invite their friends that then generates a lists of restaurants that they would like.

In addition, there would be an additional feature called Plate Off that allows users to compare each of the restaurants by swiping left or right to determine if they like the restaurant or not. There would then be a ranked list of restaurants for users to see which place they would like the most.

I fleshed these ideas out into low fidelity sketches and created some user flows to capture how users would navigate to and through the feature.

LoFis of Food Blend and Plate Off

User Flows for LoFis

Visual Inspiration and Information Hierarchy

Before creating medium fidelity designs of my idea, I explored currently existing platforms that had similar features.

Spotify’s Blend with friends.

Google Map’s various ways of filtering and giving direction.

Tinder swiping left and right express like or dislike.

OpenTable’s reviews and details about what information to display.

Through this market research, I outlined the framework of the Food Blend feature. The main structure of the feature was inspired by Spotify Blend and Tinder’s interactive design of swiping. OpenTable’s reviews and details inspired what details to include in the feature. The categories and filters were inspired by the features within Google Maps.

Beyond visual inspiration, I wanted to plan out where this feature would exist in the current app of Beli by drawing out an Information Architecture.

Information Hierarchy

The app’s purpose is to promote food discovery with a social aspect to it, and the feature accomplishes this by sharing recommendations that would be compatible for multiple users.

I decided to place the new feature under custom guides to reduce repetition of features. It follows the same theme as guides, allowing users to create a collection of restaurants while introducing a collaborative component. I had initially made Plate Off an additional filtering component which was an optional path for users to take. After receiving feedback for my TA, I decided to make it precede creating a list so that all features are being used. It would additionally lead to a more accurate, ranked list for users.

Medium-Fidelity Ideation and Iteration

After exploring different flows or ways the user could be introduced and interact with the feature, I wanted to focus on three user flows.

1. Being introduced to the feature immediately in the feed
2. Going to a specific friend’s profile to create a blend
3. Creating a new blend under the lists as a custom guide

MidFis of Entry Points

For each of the entry points, I had explored different ways to guide the users to explore this feature. In the information hierarchy, the final result of this feature would be stored in “Your Lists” but would have different ways to first interact with the feature.

Entry Point Iteration 1.0

Pros: Easy to distinguish among the lists as the custom guides are placed above the pre-made guides to encourage engagement while still following the theme of collection of restaurants.

Cons: Redundant as there are two buttons that signify creating a blend. It might be hidden as the user needs to navigate to lists and then guides.

Entry Point Iteration 2.0

Pros: The entry points is visible right away when the user enters their home page which increases engagement.

Cons: Interrupts the purpose of the feed to display the activities of their friends and might not properly reflect the relevance of the product with its placement in the information hierarchy.

Entry Point Iteration 3.0

Pros: Already chooses the person that the user wants to create the blend with. The feature hopes to encourage food exploration through a social aspect which would be beneficial if it is integrated into a friend’s profile

Cons: Repetitive of the feature “Places you both want to try” and could potentially hidden as the user navigates to each friend profile.

I decided to proceed with Iteration 3.0 as it best aligns with the goal of promoting personalized discovery. It isn’t as hidden as the other iterations as the friend’s profile can be accessed in more than one way. My TA provided feedback on Iteration 1.0 as she expressed her confusion for the difference between the “+” and creating a blend. I decided that the guides section would be a way to store all guides created rather than the entry point for them. While introducing the feature on the feed in Iteration 2.0 promotes visibility, it does not follow the overall theme of the page, disorganizing and overloading the content on one page.

After designing the entry point, I wanted to explore different ways that the user could select the friends that they wanted to share the blend with.

MidFis of Friend Selection

After exploring the different iterations on how to send the invite to multiple friends, I first decided that Iteration 4.0 would be the best decision. For the screens leading to external links, I decided it was best to omit those features since they are repetitive, but also takes the user to a different app. The goal would be to keep the user’s engagement condensed and centralized. The “Create Party” would allow the user to send the invite to multiple friends, but decided that the feature would take the user through more clicks. As a result, the Iteration 4.0 was created so that the user could only invite people who are active Beli user but can automatically choose whether they would want to send it to more than one person.

However, at this point of the project, I had to once again reassess the feasibility of this feature. Before beginning the case study, it was important for me to recognize that Beli was a relatively new app — meaning that it is still shipping new features. In this step of the design process, I failed to recognize that Beli did not have a messaging system which would make it hard to organize or send invites to a large party. With this in mind, I decided that the a friend selection was not necessary.

The next step was to explore how the user would go through the card swiping process — interaction design to make a decision.

MidFis of Card Swiping

Iteration 1.0 leads into Iteration 2.0 where the information about the restaurant displayed are the themes and their top dishes. This option was not chosen because of the redundancy of information as “View Menu” gives users to the same access. Iteration 4.0 shows the different images as smaller section of cards which is unintuitive and can be confusing to navigate as it prompts users with multiple cards they might swipe. I decided that Iteration 5.0 accurate represents the themes and stacks the images to condense them to swipe together.

The next step focused on brainstorming for the buffering screens and how to display those results.

MidFis of Buffering Screens and Results

The buffering and results are represented by multiple steps. Initially, the buffering screen was going to show the progress bar of the other person, but displays information that keeps the user idle and waiting. Rather than building anticipation, it disengages them at this point, so a small blurb is designed to replace the progress bar.

In 3.0, we explore the preview of the app and how users can choose to immediately save the blend or view all of the restaurants they have in their ranked list. If they choose to view the full blend, there is still an option for them to save this within guides.

With these design choices, I came up with three finalized flows, exploring the different entry points and how the blend gets saved through each. When creating my final medium fidelity sketches, I wanted to incorporate a page that would explain how the app is used. I chose for the results to have a preview of the restaurants the top restaurants the users would most likely enjoy and then gave them an option to see the full list of ranked restaurant. By doing this, users can quickly make a decision or see more options if they are unsatisfied with the three initial options they’re given a preview to, decreasing decision fatigue.

MidFi for Flow 1 introducing the feature through the feed

MidFi for Flow 2 introducing the feature within the friend’s profile

MidFi for Flow 3 introducing it within lists and guides

I finalized these three medium-fidelity flows to high-fidelity flows.

High-Fidelity Prototyping

After finalizing the designs for each of the flows, I prototyped them and conducted user testing. By the end of testing my prototype, I wanted to find out:

1. How did food exploration with friends feel?
2. Why did you feel that way?
3. How do the users interact with creating a Food Blend?

After interviewing 3 users, I noted some trends and insights for improving my flow and design.

With these insights, in mind I decided to:

1. Create a new button “View Menu”

In addition to adding this feature, the interaction in which the users tap through to see whether they’d like the place or not was removed. There would be one picture that represents the restaurant and the additional feature would take them to the general restaurant view where more information about the place can be found. By doing this, the user can have distinguished actions for looking for more information and inputting their preferences.

2. Removing the percentages and replacing it with which friend swiped left or right on each page

Users said that it was unclear what the percent reflected in terms of a group setting and showing their preferences. While it can mean that most of the group would enjoy the place, it over condenses everyone’s opinion into one number. I brainstormed to show either person liking the place by displaying their profile which is more specific than a number.

3. Nested under a friend’s profile, but all of them are collectively stored under guides.

The place in which Blends were stored would be under lists and guides, but they would only be made through a friend’s profile. By reducing the features present within lists, users can quickly navigate through multiple blends they are interested in and discover the feature in a more social aspect.

The Final Product

Final Prototyping

[embed]

Design Kit

Conclusion and What I learned

Based on this experience, I learned that the design experience can be messy but highly-rewarding. It was difficult since Beli had developed two new features that caused me to change my people problem and design solutions.

There were some parts that took longer than others, but had taught me all the various steps in creating a product that would serve the people. It allowed me to dive deeper in to problems with a product and how my ideas can serve as solutions to these issues.

This is a case study created for the Introduction to Digital Product Design Course. I am not affiliated with Beli.