The problem Simplefin solves

Personal finance management tools aren't built for indviduals, they target everyone. They do not gather data from multiple sources and give a holistic view of overall finances.
They throw a lot of information at the user and the user has to make sense of it. We leverage the power of machine learning to give actionable insights to the users to motive savings behavior.

What's the problem? Current apps suck at these,

• completely ignores personal context of the individual
• low financial awareness
• privacy issues with saving data
• steep learning curve in understanding markets and personal needs
• negative feedback loops
• too technical or jargon heavy

How do we fix it?

• starts with user's context
• defines baseline parameters for personal finance goals
• make a relationship with the user
• extremely clear with consent, data, and privacy
• customizable personalized widgets
• language-agnostic chatbots
• ethical marketplace for products and services with clear dislaimers
• communities for goals, groups and communities
• micro-lessons and awareness through teaching

The longterm idea is a set of apps, APIs and interfaces that come together to help create feedback loops that are tailored for each individual's personal requirements.

Challenges we ran into

Challenges with the Account Aggregator API

The biggest hurdle was trying to figure out ECDH the encryption standard that the Account Aggregator framework uses.
The workaround was to use MoneyOne API's to do all the heavy lifting.

Unstable MoneyOne API's.

Encountered two major bugs while developing the service.

The first was an API bug that prevented us from fetching the FIData.
The solution was to just mock everything possible and wait for updates from the MoneyOne Team.
The other bug was with the webhook, which just sent an empty response to our servers. The workaround was to just punch in the consent approvals manually.

Inability to test the product

The biggest let down was the stark realization that we won't be able to test the product on a wider scale since the data is all dummy and sandboxed.

Dialogflow context switches

The combinations of flows that could originate were difficult to implement in code, this involves context switches, dynamic questions, etc.
The solution was a lot of if-elses :D