PhaseWise

the spark

Framing the Problem

Traditional banking apps help users understand where their money goes, only displaying rigid transactions
These often overlook personal circumstances that can influence financial behaviour
This information does little tFraming the Problem
- Traditional banking apps help users understand where their money goes, only displaying rigid transactions
- These often overlook personal circumstances that can influence financial behaviour
- This information does little to help users prepare for expenses that they can actually anticipate
- For the purpose of this demo as well as the lack of banking data available, we generated a synthetic customer, “Jennifer”, with realistic transaction data
- For women who experience changes in energy levels, mood and other symptoms throughout their menstrual cycle, these changes may coincide with different spending patterns.
- For example, between periods of increased fatigue, a person may notice a greater unexpected spending on takeaway food
- Without preparation, these expenses may disrupt budgets, savings goals and financial wellbeing.
- The problem is that traditional budgeting tools rarely help users anticipate their individual spending behaviour before it happens
- Our goal is therefore to help women specifically understand their individual financial behaviour, prepare for predictable changes in spending and maintain control over their financial goals.
Idea Explanation
- Our idea is a cycle-aware financial forecasting app that predicts potential spending patterns, and presents them through a visual weather forecast
- Using a period tracking API and Starling’s Saving and Budgeting Features on Engine, the app combines the user’s health insights with their personal financial goals to estimate their spending habits
- Our intentions are not to say that women are unable to make sound financial decisions around their cycle, but rather to help them make use of spending patterns around the time for better budgeting and therefore saving.
Implementation
- We have a mix of traditional machine learning as well as AI integration.
- We call two APIs, a mock period tracker and mock starting engine core.
- For the frontend, we created a React native app using Expo
- It communicates with our Python FastAPI backend through REST API requests.
- When the user opens their dashboard, the frontend sends an HTTP request to the backend.
- The backend retrieves the relevant financial transactions, cycle information, AI-generated insights and returns the results in a JSON format.
- Backend in Typescript and Python
- PostgreSQL for the database
- Machine learning libraries: Numpy, Pandas, Scikit-learn and Swift charts for data visualisation.
- Dev tools: IDE Xcode, Github
Challenges
- We faced several challenges during development. Firstly, we were unable to find an open period-tracking API, so we built a mock API that simulates data from platforms such as Apple Health and Health Connect.
- Secondly, we lacked access to real banking data, so we generated synthetic bank transactions to simulate spending patterns across multiple menstrual cycles.
- We also had to ensure that our forecasts remained informative rather than overly restrictive, while avoiding unsupported claims about the relationship between menstrual cycles.
Accomplishments
- Throughout our project, we successfully developed a working prototype that combines menstrual cycle tracking with financial forecasting.
- We created a mock API for cycle data and generated 300 simulated bank transactions across six cycles to explore potential spending patterns. We implemented daily spending forecasts and phase based mood indicators to help users better understand their financial habits.
- We also designed a cashback and offers feature that connects predicted spending patterns with relevant discounts, helping users save money on purchases they may already be planning.
- Overall, we learned how to integrate health related data, machine learning concepts, and financial tools into a single user focused application, while prioritising privacy and user autonomy.
Next Steps
- Real data: Instead of using mock APIs, we should be able to connect with things like Apple Health
- Larger training dataset for ML - Train on a larger and more diverse dataset, including irregular cycles, to improve forecast accuracy and distinguish genuine patterns from random variations without assuming that cycle phases directly cause changes in mood or spending
- Expand cashback and discount services - Partner with retailers and cashback providers to offer personalised discounts based on predicted spending needs, allowing users to earn cashback on purchases they were already planning to make.
- Develop the savings feature: Introduce automated transfers into a savings buffer, helping users build financial resilience without restricting their spending.
- Pilot with a banking partner: Explore integration with services such as Starling Engine to make financial insights and savings features available through a real banking platform.
o help users prepare for expenses that they can actually anticipate

For the purpose of this demo as well as the lack of banking data available, we generated a synthetic customer, “Jennifer”, with realistic transaction data

For women who experience changes in energy levels, mood and other symptoms throughout their menstrual cycle, these changes may coincide with different spending patterns.
For example, between periods of increased fatigue, a person may notice a greater unexpected spending on takeaway food
Without preparation, these expenses may disrupt budgets, savings goals and financial wellbeing.
The problem is that traditional budgeting tools rarely help users anticipate their individual spending behaviour before it happens
Our goal is therefore to help women specifically understand their individual financial behaviour, prepare for predictable changes in spending and maintain control over their financial goals.

Idea Explanation

Our idea is a cycle-aware financial forecasting app that predicts potential spending patterns, and presents them through a visual weather forecast
Using a period tracking API and Starling’s Saving and Budgeting Features on Engine, the app combines the user’s health insights with their personal financial goals to estimate their spending habits
Our intentions are not to say that women are unable to make sound financial decisions around their cycle, but rather to help them make use of spending patterns around the time for better budgeting and therefore saving.

under the hood

Implementation

We have a mix of traditional machine learning as well as AI integration

We call two APIs, a mock period tracker and mock starting engine core.
For the frontend, we created a React native app using Expo
It communicates with our Python FastAPI backend through REST API requests.
When the user opens their dashboard, the frontend sends an HTTP request to the backend.
The backend retrieves the relevant financial transactions, cycle information, AI-generated insights and returns the results in a JSON format.
Backend in Typescript and Python
PostgreSQL for the database
Machine learning libraries: Numpy, Pandas, Scikit-learn and Swift charts for data visualisation.
Dev tools: IDE Xcode, Github

what got in the way

Challenges

We were unable to find any open period tracking APIs. To overcome this, we built a mock API that returns the same data as Apple health or any tracking app would do
Generated bank data b

high fives

Accomplishments

What did you learn? What did you accomplish?

Throughout our project, we were able to

since then

Next Steps

What are the next steps for your project? How can you improve it?

Real data: Instead of using mock APIs, we should be able to connect with things like Apple Health
Larger training dataset for ML, which can help with learning how to deal with irregular cycles and also differentiating whether a behaviour is truly due to a cycle-based mood change or something random
Incorporating discount services with cashbacks and offers