NutriFlow, healthy eating always within reach

Problem. People need to follow a diet prescribed by a nutritionist, but they struggle to prepare or order those meals, because they're personalized and made by hand.

Solution. NutriFlow is an app that turns the nutritionist's meal plan into a practical routine: the AI reads the uploaded plan (photo or PDF), organizes the meals automatically, and delivers them ready at home, helping the user stick to the prescribed diet.

That's the problem NutriFlow sets out to solve: the gap between what the nutritionist recommends and what a person actually eats day to day. I compared it to iFood on purpose, the convenience benchmark everyone already has in mind, and it makes the gap obvious: ordering junk is easy, sticking to the diet is the hard part.

The healthy-food market in Brazil is huge and still growing. According to the Ministry of Health (2024), more than 60% of Brazilians are overweight, and the healthy-eating market moves over R$ 40 billion a year in the country.

A qualitative USP study of patients who dropped out of nutritional treatment (Santiago et al., 2023, RASBRAN) found difficulty accessing food to be one of the five main categories of reasons for dropping out, alongside how strict the diet is and how hard it is to fit the plan into daily life.

I analyzed the food-delivery, meal-kit, and nutrition-software markets, looking for where those services fall short and where there was room for something new: a solution truly focused on meals personalized and adapted to medical prescriptions.

Before going into the field, I used a CSD matrix (Certainties, Suppositions, Doubts) as a filter. Instead of asking everything, it showed me where I was too confident without proof. The Suppositions became the research script, and the costliest Doubts (pricing and chef buy-in) were the first ones I went after.

Quantitative research (83 respondents via Google Forms) combined with qualitative (5 remote interviews with end users, 20 to 30 min each, over Google Meet).

Interview takeaway: people don't quit the diet for lack of willpower. They quit because, in the daily rush, they have no way to execute the plan. 4 of the 5 interviewees named iFood as their convenience benchmark.

With the persona in hand, I mapped what Ana's first week in the app would look like, phase by phase, to find where the experience lifts and where it falls apart. More than listing screens, I wanted to find the most fragile moment, the one that decides whether a person stays or gives up.

The emotion curve pointed to that moment: right at the start, when the app has to read her plan. If the AI misreads it up front, the person loses trust and rarely comes back. That's what made me treat reading the plan as the most critical part of the product, not an onboarding detail.

I started with the Customer for a simple reason: they're the one who feels the pain and pays the bill. Without validated demand on that side, it wouldn't make sense to touch the chef's or the courier's app. So this study covers only the Customer app, from uploading the plan to the order arriving at the door.

After prioritizing ideas on an impact-vs-feasibility matrix, I prototyped the 4 main flows in low fidelity: onboarding and upload, AI processing, Home (My Plan), and placing an order.

Green is the base color because the app is about food and health, and green already carries that idea without explaining. I kept orange only for what needs attention right now (the current meal, an alert), so it doesn't compete with green day to day.

Typography: Inter, the native font on iOS 17+ and Android 14+. Since people already see it every day, reading feels more familiar (Jakob's Law).

Accessibility: minimum 4.5:1 contrast (WCAG AA), 44×44pt minimum touch targets, Dynamic Type, and labels on every icon.

1. The AI reads the plan; the person types nothing. Research showed that logging food by hand is exactly what wears people out and makes them quit apps like MyFitnessPal. If the goal is to take effort out of the way, asking someone to log every meal, every day, would just repeat the problem. So they photograph the plan and the AI builds the calendar.

2. Ordering fits in three steps: period, delivery, and payment. Early versions had more steps, and every extra screen was a chance to give up. The closer it feels to ordering food, the lower the friction.

3. The Home answers a single question: “what do I eat now?”. The current meal is front and center and everything else is secondary. Fewer decisions on screen, less room to break the diet.

Moderated remote test with 5 participants (25 to 45 years old, delivery users), with a Figma prototype shared via link + Google Meet, 20 to 30 min per participant.

I used the HEART framework to avoid vanity metrics. Downloads and order counts matter, but what really counts is adherence: is the person still following the diet after a month? That's the number that tells you whether the product solves the problem or just looks nice.

NutriFlow doesn't replace the nutritionist. It just makes it possible to follow what the nutritionist has already prescribed.

The project grew out of a real problem that affects millions of Brazilians: the gap between the meal plan and daily execution. The research made it clear that what stops people isn't motivation, but the lack of time and convenience: 78% quit the diet within 3 months, and 74% point to lack of time as the main reason.

Remote tests with 5 participants showed the direction made sense: the ordering flow hit 100% success, in 58 seconds. The SUS of 78/100 points to good usability, with room to improve.