A few weeks back, I had received a call from a financial product marketer, who wanted to sell me an exceptionally good investment product and the marketer knew an approximate of what I earned, but the issue was that I wasn't able to invest due to financial stress I was going through at that point of time. The product offered was good, and they had the data on me to know that I can afford to make the investment - but they failed to understand my financial situation; Which got me thinking, if only campaign could've understood my sentiments and stress beforehand, which could've enabled them to recommend the products more effectively and would've resulted in a conversion.

What it does

FinFit is a Smart Customer Onboarding and Engagement app, that uses retail FFDC Customer Onboarding Open APIs to onboard customers using product recommendations based on customer's Financial Fitness. It takes the customer through an interactive questionnaire which measures customer's financial Stress, well-being stats, and collects customer's demographic information resulting into a FinFit score which is used to recommend relevant products and onboard into retail system with selected product.

Data is the new FUEL - With FinFit app and growing FFDC ecosystem, it could bring in very useful analytical data and patterns related to how product selection relates to customer’s financial situations during on-boarding, thus enriching onto platform's data services

FinFit is also designed to be GDPR compliant. The app takes the customer through a series of questions (10) which are divided into three segments. Customer is asked to rate the questions from 1 to 10 a. Financial Stress - Measures Financial stress the customer faces at that point b. Financial Well being - Measures if the customer is able to live the life with the current financial situation c. Financial Status - Measures his satisfaction with his credit score, credit and mortgage etc.

The customer is lead to a demographic questionnaire page, where the questions now change to his age range, we make sure to collect only the non-personally identifiable information from the customer, any personally identifiable information otherwise entered is filtered out (Area code of the postcode is requested from the customer, if the customer accidentally enters the full postcode, area code is separated out)

From the answers, FinFit comes up with a score (average of the answers given in each segment) and compares it with the historical data to understand if the score matches the other customers belonging to the same demographics and assess the difference. The calculation is conveyed to the customer and our assertions of the customer's financial status is provided, along with the products which might help the customer overcome the hurdle.

The recommendations are shown through a machine learning algorithm, based on the feedback from the others customer who belonged to the same demographics and received a similar score.

The recommendations get more accurate over time as we're dynamically building the data-set based on the incoming customer information.

Once the customer selects the most suitable product, customer gets on-boarded through FFDC onboarding API (The initial checks to see if the customer is eligible for the product has not been included for sake of simplicity, but can be added as a future addition)

How I built it

FinFit is built on the responsive JavaScript front-end framework VueJS, service layer on NodeJs, Azure SQLDB is used to store the user survey and demographics information. Azure ML Studio is used to create a linear progression ML model for score calculation and comparison. Bayesian inference recommendation algorithm is used to recommend the products to the customer. It uses Retail Banking Customer Onboarding API in FFDC to on-board customers to Finastra's core products

Challenges I ran into

Understanding and implementing ML for the first time and which ML algorithm to suits best for the role.

Accomplishments that I'm proud of

Learning ML, Implementing the ML, utilizing Azure SQL DB to continuously improve the data-set. Seeing this idea in winners list for City Hackathon

What I learned

One phone call from a random marketing person can be the beginning of a beautiful adventure :)

What's next for FinFit

  1. The eligibility check for the customer, before the customer gets on-boarded through FFDC Customer Onboarding API.
  2. Integration with Retail Product APIs to filter product recommendations list
  3. Bring in very useful analytical data and patterns on to how product selection relates to customer’s financial situations during on boarding. DATA is the new FUEL !

Built With

Share this project: