Yum-Me

Food Recommendation App that provides recommendations through the reviews of users who share similar taste preferences. 

Role: Product Owner and Designer

Duration: 3 weeks

Tools: Figma, User Research, Case Study 

Context: Passion project

Have you been disappointed by food reviews?

Well, I have.

I know how it feels to have seen raved reviews of a restaurant, went to try their food, and it was not what you had expected.

I know the feeling so well that I drew a comic about it once: 

I did a quick poll on Instagram. 

85%
of the respondents have been disappointed by food reviews as well. 





Ah, that's Singlish...

But... why do we get disappointed?

I love food, I love to eat, and I love good restaurants.
Reviews are meant to help consumers make smart decisions on where to eat, and not misguide us. Surely, people aren't lying in their reviews — right? 

So I ran a case study to investigate.  

Case Study

I chose Tipo Pasta Bar, an Italian restaurant in Singapore famous for its variety of pasta. It was particularly popular at that time as it allowed restaurant-goers to customize their pasta. 

Finding

Curiously, I found a wide mismatch in reviews of Tipo Pasta Bar's pasta texture. 

It's not uncommon to find discrepancies in reviews on restaurant pages, but these mismatches in reviews can be misleading for consumers who rely on high star ratings (over 4 out of 5 stars) to judge a restaurant's quality.

Insights 

In a fast-paced world with myriad options, time spent browsing through reviews to make a decision is costly and could potentially be a waste of time.

Additionally, detailed reviews are often wordy, and difficult to digest, and users may not have the time or patience to read through the entire review.

Opportunities

With so many opinions online, how do we differentiate what is credible or not?

How can users minimize effort (efficiency) in making a food decision that is accurate (effective) to their preferences? 

How can we minimize the risk of an unwise choice?

Yum-Me:
Let your food preference decide your dinner place.

Yum-Me is a Food Recommendation Application that provides restaurant recommendations through users who share similar taste preferences as the primary user. 

It learns about an individual's food preferences through a quick survey and adopts a machine-learning algorithm to recommend restaurants that other users with similar preferences have enjoyed based on the user rating. 

The higher the rating of another user, the closer their preferences are to you. 

Predictive Model 

Through the onboarding survey, users will have to input data on the dependent variables (i.e., taste, texture, dietary preference, etc). Missing values will not be possible as the questions cannot be skipped or else the survey is considered incomplete. This is to ensure that the model is not skewed by any outlier or fluctuations in the data. 


Supervised ML algorithm

This application will adopt a supervised machine-learning algorithm trained upon features with an output value to predict the target variable; which is the similarity in taste preferences between the primary user and other users. 

The model will use a recommendation system such as the Bayesian inference algorithm to predict the similarity between the user’s preferences using new data (e.g., if a user visits new restaurants) and prior knowledge of the user’s preferences (i.e., the survey responses). 

I have temporarily put this project on hold given the many other projects ongoing in my life at this time. If you are as passionate about finding good restaurants as I am or would like to learn more about this project, let's chat!