DATA SCIENCE & FOOD

Hunting and gathering.


People with food-limited diets feel restricted in their lives. This app helps people explore and discover new foods within specific diets, with a focus on trust and delight. 

OVERVIEW        RESEARCH        STRATEGY        DESIGN        PROTOTYPE        LAUNCH

 

campaign1

overview

NEW YORK

“I'll never get to eat tasty food again.”

COMPANY

An ethical startup that focused on helping people with allergies and food limitations discover natural and delicious foods in their local grocery stores.

The food+tech startup offered a mobile app that let people filter their grocery store by ingredient, allergens, applicable diet, and flavor. Ingredient1 joined Food + Future, a collab between Target, MIT Media Lab, Ideo, & Intel.

PROJECT

This case study covers one of Ingredient1's earliest prototypes. Feedback from this experiment focused the startup’s mission by clarifying our true audience and understanding their core needs. This prototype formed Ingredient1's food and nutrition modeling.

During the startup's early experimentation and growth, we focused on people with food limitations in the New York area, and were able to partner with speciality stores and local brands in order to best provide users with information about familiar and local foods.

grocerystore2

ROLE

Co-founder, Head of Product —  Balanced user needs with startup resources, and directed vision that guided team-building, research, design, and operations. Grew from a two-person team with an idea into a mission-driven startup with a commitment to clean food and data. 

TEAM

The Ingredient1 Team included a head of business development, head of product, a community manager, data team, project manager, two developers, and five certified nutritionists.

INSIGHT

One in three people walk into a grocery store with at least one food restriction.

research

DISCOVERY

Research & Synthesis

This project's initial research centered on understanding market fit and testing assumptions before committing to the design and development of a full native app. The riskiest mistake we could make at this stage would be to misunderstand the user’s core need, and to incorrectly analyze how they buy groceries. Fitting into a user’s existing daily flow, and being effective, was critical for adoption and success.

people-1

WHO NEEDS THIS

Gathered background data on people with food allergies, and people who are on on philosophical, medical, and lifestyle diets: and found that one in three people walk into a grocery store with at least one food restriction.

  • Define core users and their needs
  • Understand their biggest pains around food
  • Analyze how that affects their food habits (shopping, cooking, eating)

 

METHODOLOGIES

  • Market research
  • User interviews
    Observational field studies.

Additional avenues of research for this project include behavioral and emotional responses to food limitations, grocery shopping mental models, allergens and FDA limits, and requirements for organic certifications.

PROBLEM

"People with allergies give thanks after a meal."

unsplash-image

USER INTERVIEWS

For this prototype, we spent two weeks interviewing 40 grocery shoppers in Manhattan with an allergy or food limitation.

They provided us with their grocery lists, weekly shopping habits, major likes and dislikes around food, and how having an allergy or limitation affects how they shop, where they shop, and what they buy.

We analyzed the grocery lists of 32 users over the duration of a month, and after eliminating non-food items, discovered that a little over 65% of their food groceries were repeated from one trip to the next. Further investigation found the various “whys” behind this: from being in a comfortable routine to fear of trying new foods.

000_1

OBSERVATIONAL

Our most successful research approach was during observational field studies, where we shadowed 15 grocery shoppers with food limitations, and asked them to communicate their decision-making process from entering the store through the checkout process.

Each trip took around 45 minutes, participants purchased an average of 20 items, and 47% of participants accessed the internet while shopping.

unsplash-image

PATTERNS

Interviews and observational walk-throughs revealed multiple mental paths around grocery shopping that can be refined into two major decision models: "Retrieve", and "Explore".

People with food-limited diets tend to “define and retrieve” as their primary grocery shopping behavior. However, this isn’t because they prefer this method, but because the necessity of reading labels and researching brands pushes them into that decision model.

Target users for this service are people with at least one food restriction, for health or philosophical reasons, who grocery shop for themselves and/or others. They prefer retail grocery shopping to online because of the ability to read labels, see products, and get help from a clerk.

INSIGHT

Food-limited grocery shoppers have more anxiety around shopping, and they spend more time spent reading and comparing labels.

USERS

Frustrations & Fears

USER PROBLEM

With over 35,000 products and at least 100 unique data points per product (including ingredients, certifications, nutrition details, brand, and price), how long would it take to read every label in a grocery store?

At a generous two minutes per label, it would take a continuous 48 days to read every product in the store.  People on food-limited diets can't opt-out of reading labels at the grocery store, which results in hours of extra time and research.


ONLINE OR RETAIL

We compared online grocery shopping with retail grocery shopping to helps us refine demographic focus. Online grocery shopping can suffer from limited stock, and missing product information like ingredient lists.

Retail shopping allows users to get immediate and effective help from a clerk, it allows for impulse purchases, and contributes to a more immersive, tactile browsing experience: something people rated as important to them when grocery shopping.

SYNTHESIS

User's Information Journey

Researching what foods a person will allergies or on a limited diet can and can't eat is the biggest pain point people experience when they are dining out, or grocery shopping.  Much of this has to do with the overall lack of information available, and the dispersed nature of where that information is located.

People with food-limited diets search online at brand websites and forums. They read packaging labels and keep their own comparison spreadsheets. They download health and nutrition apps to find and collect information, only to find it outdated and lacking critical information like allergen warnings.

painpoints

SYNTHESIS

User's Emotional States

Feelings of restriction around food, something foundational to our well-being, can affect emotional resilience as they highlight the lack of core needs around sustenance and physical safety.

People with allergies can feel like they are denied basic subsistence, and they feel like they are missing out on broader social participation and community support. It can also affect a person’s sense of identity and freedom. This contributes to situations where people who are already food-inhibited further constrict what they buy and fall into a food feedback loop that is restrictive and uninspiring.

  • Feels restricted
  • Loss of autonomy
  • “I may never get to eat anything tasty again.”
  • Desire to feel they have “discovered” something
  • Desire to feel control and self-empowerment
fundamental

SYNTHESIS

User's Behavioral Journey

Personal success with managing an allergy or diet is strongly motivated by how much acceptance of their food limitation a person has developed.

For example, someone who has eaten eggs all their life, only to develop an allergic reaction to eggs as an adult (late-onset allergies are common) will initially feel in much denial about it. That may result in them choosing to eat eggs aanyway, or be wary of substitutes because it will mean accepting their loss of eggs.

acceptance

SYNTHESIS

Meet Alice

On Wednesdays (or sometimes Thursdays), Alice goes to her second-closest grocery store after work because it has a more expansive selection of dairy and nut-free foods. She was rushed this morning and didn’t have time to make a list, so she is trying to remember items and putting a rough list together on her way to the store, and hoping she doesn’t forget anything important.

dayinlife

strategy

STRATEGY

Strategy & Constraints

EXPERIENCE GOALS

  1. Provide an experience that is minimal and effective. Allow users to quickly find what they need, and make sure large amounts of data are easy to skim and understand

  2. Provide an experience that is joyful. Help users feel curious about what they can eat, and let them explore potential snacks and meals in an obligation-free way.

  3. Provide an experience that is kind. Take into consideration the elevated anxiety of users and lean towards over-communication and showing sources.

CONSTRAINTS & LIMITS

Third-party data sources and APIs did not contain the level of data needed. In order to get the data we needed through APIs, we would need to merge multiple sources together and spend notable time managing duplicates and conflicts.

This constraint prompted us to focus on collecting the data we needed from offline sources, and began our journey of collect and organizing over a million food data points.

The deciding factor for creating own data was the research showing people want data they can trust, and that meant figuring out a way oversee and direct data integrity and transparency.

design

IMPLEMENTATION

Design Explorations

exploration2

NATURAL LANGUAGE

This set of screens shows a user flow from app launch to product detail screen. For the focus of this case study, I have omitted strategy and screens for the onboarding phase.

In this exploration, we focused exclusively on the amorphous nature of the "explore & discover" shopping model by bringing into the flow some natural language prompts that mirror the internal decision tree of a user who isn't sure what they want to eat.

exploration1

MATRIX INTERACTION

In this matrix interaction model, we minimized the decison-making into a single choice along an axis. In this flow, a user only needs to make one selection among four options. This exploration also reduced the cognitive process of filling in the blanks of a sentence. With the natural language flow, a user had to make two explicit decisions before getting a result.

IMPLEMENTATION

Prototype

Group-2

PROTOTYPE LAUNCH

For the prototype, the matrix selection model was the best fit due to its simplicity and effectiveness. The prototype provided test participants with a larger call-to-action for the "Surprise Me" button, and the ability to favorite a product. After finalizing the remaining screens, and going through smaller design iterations, we were ready to test this experiment.

Ingredient1 partnered with two natural grocery stores in Manhattan to share the prototype with their in-store shoppers. This provided us with an potential number of daily shoppers ranging from 200 - 750.  Of those, we were able to attract 5% of shoppers.

LOCATION  —  NY, NY
DURATION  —  2 months
PLATFORM  —  Mobile web

Prototype-Launch-1

PROTOTYPE LAUNCH

1. Provide for multiple mental models. 
Offered users the ability to choose between “search and retrieve” or “explore and discover” to help us understand further nuances about decision-making.

2. Design a clear flow for selecting between search/surprise options.
Matrix interaction model limited misclicks from confusion with the UI and allowed us to focus on the patterns around decision-making like “savory snack” or “sweet snack”

3. Single screen about each product.
Provided detailed information for each product to measure usefulness to user, preference in order of information, and preliminary screen drop-off risks.

4. Ability to favorite a product.
Tapping on a heart icon allows the user to complete the discovery loop by making a decision, and gave us supplemental data on how to best implement this feature. Crucial to completing the discovery loop: imagine, explore, decide. 

5. Functional keyword search results. 
Though prototype database was limited, all the product information within was complete and robust, including basic dietary and allergen tags, ingredient list, and nutrition information.

PROTOTYPE

Outcomes & Learnings

After reviewing app metrics data and talking to prototype participants, the results of the prototype provided us with valuable insights into improving the feature and overall experience. Users wanted more options and wanted to make fewer decisions to get there.

nav

NAVIGATION MODEL

Our testing criteria compared which shopping model users choose. Was "Search" selected more often than "Surprise"? And how did that balance out on repeat visits? We couldn't find anything definitive around which model users preferred. We did find the "Surprise" feature to be an effective funnel into searching, and one of the first things users do on repeat visits to the prototype.

This helped us determine final hierarchy for the full native app, placing "Search" and "Discover" parallel to each other on the main level.

options-1

UNIQUE OPTIONS

Based on user feedback, we added unique options like "crunchy", "local" and "organic" to the discovery critera. For each option offered to users, we updated each product in the database and analyze whether they meet the critera for that variable.

random

DECISION FATIGUE

In the prototype a user can choose one among four options (sweet snack, salty snack, etc) to get a surprise food result that fits those parameters. Feedback from test participants said they didn't want to decide among the options every time they wanted to discover a new food. Sometimes they didn't know what they wanted to the extent they didn't know, or care, which option was selected.

brandverified2

TRUSTHWORTHY RESULTS

Above all, users wanted to know they could trust the accuracy of the information they get about each product: allergens, applicable diets, nutrition information, and certifications.

For this, we created a three-step data verification process, that includes review by certified nutritionists, for onboarding our food brands. Once completed, products in the app displayed a "brand verified" banner.

dev

PRODUCT

Feature Release

This prototype iterated into a discovery feature, and was included as a cornerstone of the full native app, alongside the ability to search foods, save favorites, find store inventory, and manage allergen and ingredient preferences.

The app launched with 2,500 users, partnering with 5 NY grocery stores, and containing detailed information and analysis of over 35,000 natural food products.

discoveryapp2
Final Feature Flow2
participating_brands

BUSINESS IMPACT

Because we focused heavily on transparency and accurate data, the best way for a food brand to increase their product's visibility in the app was to increase their product's data integrity.

With 55% of our users playing with the discovery feature at least once per visit, and 49% of of those users more than once per visit, it became the feature that drove engagement to other areas of the app, like search and favorites. This allowed us to work with our partner food brands to increase data integrity by verifying their product information.

When a natural food brand partnered with Ingredient1, the brand verified their product information through a client dashboard. Direct ingredient and nutrition verification by the food brand allowed those products to show up in Discovery feature results. A "brand verified" badge appeared on each applicable product's detail page, and showed up in search results.

discoverflow1

PRODUCT CURATION

Having relevant discovery results for users was a major feature goal, and resulted in flagging a subset of the products in our food database as "discovery-ready", foods that contained complete and verified information. We focused on including a variety of foods that fit within each diet and allergen variation we supported in the app.

For example, if you told the app you had an egg and wheat allergy, you were vegan, and you wanted something creamy and sweet... we made sure there were at least a dozen results available through the curated discovery feature.

userprefs

SWEET OR SPICY?

After assessing detailed metrics from the discovery feature that focused on user-selected options, we found the top three most popular criteria, and the three least selected options to discover foods.

Users' top three  favorite options for discovering foods were "Small", "Sweet" and "Savory". 

People were least interested in choosing foods that were "Large", "GMO free", and "Chewy". We later removed the GMO option and replaced it with "Local" and "Small Batch" based on user and brand feedback.

Flavor was a higher factor than texture, though "Creamy" as an option was selected more often than "Crunchy", except when paired with "Spicy". 

launch

THANK YOU

Hunting and gathering.

CONTACT
ERIS@ERISFREE.COM

 

© ERIS STASSI 2020