Foodflow

A mobile and desktop app created in twelve weeks to support Milanese food banks and NGOs in managing their resources more effectively.

Skip to

STAR & Process

Problem Definition

Chosen Concept

Prototype & Test

Final Solution

Lessons Learnt


01 SITUATION

This course primarily focused on research. We were assigned a topic and tasked with identifying challenges and opportunities for technological innovation:

COVID-19 increased poverty in Milan. Even some working families have started to need food aid, creating the new category called "working poor".


02 TASKS

  • Identifying the right problem;
  • Defining a concept;
  • Improve the concept through multiple research methods and cycles;
  • Choose a final concept;
  • Prototype and test with users;
  • Design the final solution.

03 ACTIONS

  • Supported research and concept creation;
  • Worked with a data scientist to train a predictive model;
  • Facilitated 2 out of 4 Design Reviews to test the final concept;
  • Designed multiple prototypes and user tests;
  • Design System & UI.

04 RESULTS

If implemented, the app could cut inventory time in half by eliminating duplicate checks. Currently, food banks inspect donations before sending them to NGOs, who then recheck the items upon arrival. Testing demonstrated that the app would significantly expedite this process for both parties.

Problem Definition

background

Defining the Final Concept

SELECTED PROBLEM

After extensive research we closed on the following problem to tackle:

Food aid organizations have been experiencing long lines and staff shortage.

SELECTED CONCEPT

With the chosen problem and our research insights in mind, we shifted to a third concept, concentrating solely on aid organizations:

An app to streamline food storage management, enabling aid organizations to distribute food donations more efficiently by prioritizing items based on their expiration date and nutritional value.

Concept's Storyboard

Images generated using DALL-E.

Research questions:

Prototype & Test

background

Final Solution

After several iterations, we established the final information architecture and crafted the desktop interface (tailored for Food Bank & NGO managers) and the mobile app (tailored for volunteers).

Information Architecture

Ontology

Taxonomy (Cupboard feature)

To check the rest of the taxonomy, click here.

Choreography

Design System

I designed a design system part of the final screens for the mobile app, while another colleague designed the ones to the desktop app.

background

Impact

Impact

Accessibility Considerations

AFFORDABLE ACCESS

Small NGOs can use the app at no cost. Medium and Large NGOs can upgrade to a premium version with advanced inventory features for $300, significantly less than the rate ($1,357) some of them pay for similar systems.

OPTIMIZED FOR COLOR VISION

The app screens were checked for color blindness accessibility (protanopia and deuteranopia types) using Photoshop's proofing feature.

VISUAL IMPAIRMENT

While images have alternative text descriptions, interactive graphs remain a challenge. The current solution is a data export feature, since tables are more accessible for screen readers. However, more engaging solutions should be explored.

Lessons Learnt

Addressing our cognitive biases earlier could have saved time.

We had assumptions about NGOs needing more donors that led to time-consuming errors. Discovering system waste was surprising. In future projects, I'd include a moment to map the team's assumptions, classify the riskiest ones, and include them in the research plan.

What people say often differs from their actions.

Initially, we used interviews and surveys to address RQ2 and RQ3 but couldn't find clear opportunities. Switching from attitudinal to behavioral methods provided tangible insights. This gave me a deeper understanding of when to resort to attitudinal and behavioral methods.

Machine Learning has a different iteration pace.

ML iterations are highly influenced by data dependencies, uncertainties in model architecture, and specific performance metrics. Unlike the deterministic feedback in software development, ML requires more exploratory cycles. Deploying ML models also presents unique challenges, such as model drift and retraining.

Team

Aylin Paçaci

Beatrice Fidone

Carolina Iplinsky

Giovanna Mundstock

Juliana Maines

Jun Deng

❤️ Special Thanks to:

All food banks, NGOs and local business that participated.

My team, for their unwavering commitment to this project.

My professors, for their valuable guidance.

Project developed in the course of Final Synthesis Studio from Politecnico di Milano.