AI-Powered food logging to track health trends over calories

Vively

AI-Powered food logging to track health trends over calories

Led the redesign of food logging to focus on frictionless food logs in seconds, powered by AI with a focus on health trends over merely just tracking calories.

FOCUS

AI Feature Development

AREA OF WORK

Healthcare

ROLE

Lead Product Designer

PROBLEM

Health Trends are more important than Calories

People forget what they eat. They forget what they had for dinner, they forget the little snack they had between brekkie and lunch, was it 3 or 4 eggs for breakfast yesterday? Tracking has become an integral part of the health-conscious space, with people tracking their daily food intake, macros and sometimes even micronutrient intake. With Vively's food tracking, we wanted to provide users with an additional avenue of tracking their health metrics in a way that's meaningful, and more importantly, simple.

1

Food logging is tedious. For a lot of users, breaking down a meal into its individual ingredients is confusing and often a process that takes time. Many users may outright miss out on ingredients or under/overestimate the weights of foods.

1

2

Numbers don't mean much. A calorie, protein or vitamin doesn't mean much to most users. Numbers are useful, but without explicit guidance or actionable feedback from the app, the user can't gain a whole lot of value from pure numbers.

2

DESIGN APPROACH

Analysing mainstream design patterns and aligning design with business requirements

We're not reinventing the wheel. Food logging is already a well-established design pattern across a number of apps so the approach to save excess time and resources being poured into the feature design was simple - To build a food logger that performs simply and feels natural, and a food logger than integrates nicely into the Vively ecosystem.

1

Analyse existing design patterns.

From MyFitnessPal to Cronometer, there's an abundance of food tracking apps, each with their specific gimmicks or feature-sets, but at their core - They log and track food. We ran a competitive analysis across a number of apps under a basic SWOT analysis each to determine where we should draw focus on in design while aligning with the business goals.

1

Analyse existing design patterns.

2

Design and iterate with a UX Object Map.

Using a similar process to the first design approach, we then started building low to med fidelity designs, assess under a basic SWOT, and then build the next iteration. To measure against a foundation, we established a UX Object Map with the micro-features and parts required of the macro-feature to ensure the designs are meeting some sense of requirements.

2

Design and iterate with a UX Object Map.

3

Ensuring AI provides value.

AI is a tricky path to navigate when integrating into any feature. Designing an AI feature requires a number of guidelines to be followed. These guidelines are a preference and not based off any existing foundation. These include: The AI output must make sense to the user, the AI output must save meaningful time or effort, the output must be forgivable, the output must be reviewable or editable, and the

3

Ensuring AI provides value.

DESIGN DECISIONS

Putting ease at the forefront

Building something simple was always at the forefront of the design approach, but simple could mean a number of things. It could mean clutter-free, avoiding excessive visual flair, reducing the amount of clicks to achieve something. But simple here had to meet a certain criteria; It had to be easy enough for an older demographic of users to use, something that required a little amount of actionable steps to log a meal, but also providing the flexibility power users expect from an app to fine tune their results, and more accurately log their food.

1

Simplicity as the default. The default action always when logging food was having the camera open. Instead of switching to the camera, one less action meant users can instantly capture a meal in front of them, without ever needed to switch to a new view. For power users, manual logging was hidden by one extra press, but visually tied to the camera view so they know switching modes is always a forgiving, easy task.

1

2

Focus on the food, not the calories. We wanted users to focus on accurately capturing and logging their food, without the burden of worrying about their calories or macros. For this, calorie and macro details were pushed to a secondary action input and not a primary visual cue. Users can comfortably log their meals and then be presented with the health, calorie and macro breakdown afterwards.

2

OUTCOMES

Measuring feature success

Although in our subsequent user testings, we did find some problematic UX decisions, overall, the feature proved successful. We observed measurable improvements in key metrics and had a number of data points improve when compared to the previous version.

16%

Increase in image attachments

Users were attaching more meal captures to their logs, driven by the new AI logging feature

11%

Reduction in manual food logs

Although this might seem like a lower retention score, the bigger picture is more important. Food logs were consistent over a 3-month cohort, yet saw a reduction in manual logging and increase in attachment logging, meaning people were opting more to AI capture their meals than manually log them.

LEARNINGS

Dog fooding might be king

We discovered a method of usability testing that yielded the best results even beyond user interviews - Dog fooding. An odd name sure, but a method that ethically, every designer needs to be aware of. It establishes a unique connection between the user and designer that can't be achieved through sheer metrics and secondary research.

1

Dog fooding helped us find bugs, UX issues and built out SWOT analyses before the feature went live, or even as the feature went live during our fine tuning and iterative sprints.

1

Dog fooding helped us find bugs, UX issues and built out SWOT analyses before the feature went live, or even as the feature went live during our fine tuning and iterative sprints.

2

More written UX documentation should've been established over diagrams and workshops, creating one central source of truth for documentation.

2

More written UX documentation should've been established over diagrams and workshops, creating one central source of truth for documentation.

3

We could've taken further advantage of the Vively ecosystem and built in features had it not been for food logging being worked on much earlier on. The tight integration of features and ecosystem could've proved food logging to be even more enticing to users, and bring more value to the existing feature-set the app had to offer.