Explore

Upstart Case Study - A 'Breezi' Story

Making applying for a loan a breeze.

Arif Islam

Project: LoneStar

TL;DR: Making the application for loans less cumbersome through an easy breezy digital experience

Status: Discovery

Shipping: TBD

⁠

Problem Alignment

The Problem

One of the business challenges that Upstart (and any lender) faces is validating the identity of those that apply for a loan. This verification ensures that the borrower is not using a false identity to fraudulently receive funds. Today, Upstart has a team of people that review documents provided by the borrower to verify their identity. A few problem areas that we are looking to solve for:

Improving the manual review process at Upstart, so that the manual review is minimal to none

Verification process is taking too long for borrowers resulting in drop-off or churn

Validating the identify for those that apply causes borrowers to churn due not wanting to share personal or financial information

Borrowers want/need money and expect to be approved immediately, when comparing to other lenders

High-level Approach

Hypothesis:

If Upstart could easily identify borrowers during the identification process, it would greatly increase the amount of verifications in a shorter amount of time, resulting in less drop-offs and more approvals.

Assumptions:

Loan Interest rates as low as 2.49%

No fees or costs up front for the borrower

Mortgage rates set for current market price (2.99%)

Co-borrowers are not applicable

Only unsecure loans are applicable

Allow for basis points to be applied for all types of loans

Upstart has an existing partnership with Redfin to pilot Breezi

MALI is a new cryptocurrency that beats SOLANA in terms of speed but doesn’t lose the security it needs to facilitate legal digital financial transactions. It is open-source has can be easily leveraged within mobile or web applications for financial transactions.

Proposed Approach & Business Case

For both borrowers and Upstart’s internal teams, we will introduce a digital platform called Breezi that allows for both the borrower and Upstart team to identify and quickly approve or disapprove a borrower’s request for a loan. This will be a two-sided experience that leverages AI/ML engines in order to pre-qualify candidates as well as highlight metrics for approval teams at Upstart to make quick and accurate decisions on whom to approve outside of pre-qualified candidates. Finally, in order to differentiate from other lender’s like Better or LendingTree, we will also test a pilot in partnership with Redfin and Automobile companies. Additionally, Breezi will introduce an experience that will allow borrowers to upload necessary documentation (e.g. driver’s license), where Breezi will leverage image classification to take information from the license and instantly verify the identify of the borrower.

For the borrower, an individual will provide a certain amount of information, in which Upstart will collect and consume. A soft credit check will perform as well as analysis of spending habits to return a ‘pre-qualified’ approval. Once an approval is made, Upstart will collect the borrower’s bank information to deposit the account. In short, the borrow should be able to experience the following:

Ability to provide personal information

Ability to provide SSN for a soft background check

Ability to upload/scan a valid ID

Ability to choose what type of loan to apply for

Ability to provide bank information once an approval is granted for a direct deposit

Ability to receive official approval letters through email and key notifications

*Ability to pay for a charged service fee, only if approved

For the Upstart team, these users will be receiving all applicants’ data, regardless of immediate approval. They will be able to view insights powered by an ML engine that synthesizes spending habits, consumer’s personal data, and credit history to provide risk assessments that enables faster decision making for the team. In short, the Upstart team should be able to experience the following:

Ability to view borrow information

Ability to view ‘Risk Score’s and assessment

Ability to search for applicants

Ability to receive notifications for pending approved borrowers

Ability to approve or reject a borrower’s applicants

View a consumer’s DTI (Debt to income ratio)

Possible loan programs specific to the consumer’s inquiry and circumstance (e.g. first time home buyer program)

View Basis Points loans for Fixed and Adjustable rates

For Partners (e.g. Redfin, Mercedes etc.) , with whom we have a shared service agreement, will have access to Breezi. The use case here is so that when consumers come in for an in-person experience (e.g. viewing a potential home to purchase), that staff can easily leverage Breezi’s platform to quickly approve borrowers for money and move to the next step in purchasing the asset. In short, staff from our authorized partners should be able to:

Let a consumer use Breezi’s borrower experience

Review limited approval or pending information of a borrower’s application

Present options of different rates for the borrower to select

Goals & Success

In terms of what success looks like for the above proposed solution and problem alignment, it’s important to note which metrics will attribute to the success of our solution. For both the borrower and Upstart team’s experiences, there will be separate success metrics for both. Those are:

Borrower Success Metrics

Borrower Success Metrics

Metric

Description

Metric Area

Why is it important?

Borrower NPS

This will measure our loyalty and brand of the product/service we are offering

Net Promoter Score, Satisfaction with the Product

We can segment on protractors, detractors, and passives into different areas that will help us identify improvement areas

Monthly avg. # instant approvals

This will measure the average number of instant approvals per visitor to see how many instant approvals

Fast Approvals

This will help us see how many borrowers on average are instant approved

MRR

How much money we charge for this service per applicant on a monthly basis

Revenue

This will calculate how much revenue we are generating for the product we are providing per transaction

Monthly avg. # pending approvals

This will measure the average number of instant pending approvals per visitor to see how many instant approvals

Fast Approvals

This will help us see which loans are being applied to the most and which are generating the most revenue

Avg. time to complete application

Measure the average time it takes for the borrower to complete the application

Usability

This is a key metric to see how easy it is for a borrower to submit their information

Avg. time for instant approvals

Measures avg. time for instant approvals to take affect per borrower

Fast Approvals

This effectively measures the average time for approvals to be granted once an application starts

There are no rows in this table

⁠

rifi

Upstart Team Success Metrics

Upstart Team Success Metrics

Metric

Description

Problem Area

Why is it important?

Avg. review time for applications

Measures the average time it takes to review non prequalified applications per borrower

Team Efficiency

This will help us see if Breezi helps speed up the process

Monthly avg. of non-prequalified applications that are approved by team

Measures avg. monthly applications that are manually approved by the team

Team Efficiency

This will help see how many applications are being reviewed manually vs automated by Breezi

OSAT

Overall satisfaction with Breezi to analyze and review applications/process

Product Delight

Let’s us know the team is satisfied with Breezi

WAU

This will measure Weekly Active Users with the team

User Engagement

This will help us pinpoint drop-offs or spikes of engagement for the Upstart Team

There are no rows in this table

⁠

Solution Alignment

Who Would Be Using Breezi?

Potential Borrowers

Looking for a new home

Looking for purchase a vehicle

Looking for a personal loan

Upstart Team

Real Estate Agents (Redfin)

Future state: Car Salesmen (Automobile dealerships)

⁠

The Borrower Experience

For the provider experience we want to focus on a few key areas:

Faster Approvals for prequalified candidates

Increased Conversions for when a borrow has officially agreed to borrow money from Upstart

Brand & Loyalty so that potential borrowers are aware of Upstarts value and delightful experience, especially against competitors

⭐For the sake of this exercise, let’s assume the delivery team + cross-functional stakeholders met to size/discuss each feature in order to assign the RICE score ⭐

Borrower Features (Epics)

Borrower Features (Epics)

Feature

Description

Problem Area

RICE Score

Loan Selection

As a borrower, I’d like to be able to apply for a loan specific to a type of loan, so that I can borrow money from Upstart specific to my needs

Fast Approvals

10000

Provide Income and Personal Information

As a borrower, I’d like to provide income and personal information so that Upstart can verify my identity

Fast Approvals

10000

Instant Approvals

As a borrower, I’d like to get a instant approval, so that I can access loan money faster

Fast Approvals

10000

Provide Bank Information

As a borrower, I’d like to provide my bank information to Upstart, so that I can receive a direct deposit once my loan is approved

Fast Approvals

10000

Leverage MALI crypto

As Breezi, I’d like to leverage MALI’s cryptocurrency infrastructure, so I can process borrower transactions faster and get money into their accounts with haste

Fast Approvals

9000

Receive notifications

As a borrower, I should receive notifications from Breezi, so that I can be aware of important updates

Borrower Engagement

9000

Upload valid ID

As a borrower, I should be able to upload a valid ID, so that Breezi can verify my identify faster

Fast Approvals

10000

⁠

Veriff⁠

⁠

As Upstart, I’d like to integrate Veriff into Breezi, to allow for faster verification of borrowers

Fast Approvals

9000

There are no rows in this table

⁠

The Upstart Team Experience

For the Upstart Team experience, we will want to focus on a few key areas:

Improving application processing times for borrowers

Assess pending applications for borrowers

Making recommendations to potential borrowers who have pending applications

⭐For the sake of this exercise, let’s assume the delivery team + cross-functional stakeholders met to size/discuss each feature in order to assign the RICE score ⭐

Upstart Approval Team Epics

Upstart Approval Team Epics

Feature

Description

Problem Area

RICE Score

As an approver, I’d like to be able to log into Breeze, so I can view applications that need approval

10000

View Pending Applications

As an approver, I’d like to be able to view pending applications, so that I can manually approve/reject them

Borrower Application Processing

10000

Search for Applications

As an approver, I’d like to be able to search for borrowers, so that I can quickly find borrowers I am looking for

Borrower Application Processing

10000

Review Risky Applicants

As an approver, I’d like to see borrowers who have been tagged as ‘risky’ and reasons why, so that I can make effective decisions quickly

Borrower Assessment

9000

Approve or Reject Borrowers

As an approver, I’d like to approve or reject borrowers, so that I can instantaneously let them know if their application status

Borrower Application Processing

9000

Borrower Analytics and Scores

As an approver, I’d like to see analytics and scores based in inputs and historical data, so that I can make effective decisions for pending applications

Borrower Assessment

9000

Notifications

As a borrower, I’d like to see important notifications from Breezi, so I know when to be informed and/or take action

Engagement

9000

There are no rows in this table

⁠

The Partner Experience

For the Partner experience, we will want to focus on a few key areas:

Allow for possible customers to quickly get approved so they can take the appropriate next action

Instantaneously allow for borrowers to see different options

Severely limit the amount of input for a partner

⭐For the sake of this exercise, let’s assume the delivery team + cross-functional stakeholders met to size/discuss each feature in order to assign the RICE score ⭐

Partner Features (Epics)

Partner Features (Epics)

Feature

Description

Problem Area

Rice Score

As a partner, I’d like to access the Breezi Partner application, so that I can allow consumers to give input

10000

Review Options to Present to Borrowers

As a partner, I’d like to present different finanial loan options to buyers, so that I can easily discuss and convert them to purchase

Borrower Conversion

10000

Quick Approval

As a partner, I’d like for borrowers to be instantly approved, so that they can put up the necessary upfront costs

Borrower Conversion

10000

There are no rows in this table

⁠

High-level Diagram Of Data Input and Process

⁠

Sample Mocks

⭐For the sake of this exercise, the screens below were created from Figma, just to give an idea

Upstart Team App :

⁠

Partner Experience:

⁠

Product Strategy & Monetization

Who Will We Sell Our Product To?

Though our product will service as a B2C and a B2B product, we will primarily sell out product to the following ‘businesses’ and consumers:

Borrowers

⭐Value Prop: we offer the fastest and most secure transactions through the MALI cryptocurrency network, where prequalified borrowers can get approved in seconds.

Partners

⭐Value Prop: through the Breezi borrower experience, we offer a service to home loaners, automobile dealerships, and more to offer flexible purchase options and fast on-site prequalified approvals.

How Does Breezi Make Up For Costs?

Costs & Revenue

Known Costs

Dev build time for updating training model with new inputs and adjustments to algorithms

Web/Backend build time for building out client-facing applications and consuming ML model

DEV time to incorporate MALI cryptocurrency transactional systems

Social Media Campaigning and Advertisement costs (GTM costs)

UX/UI resourcing throughout Ideation > Execution

Product Support

Revenue Streams:

At a high-level we would sell Breezi via a service level agreement or contract agreement with partners. We will also charge a service fee per consumer, for the key value prop of providing instantaneous financial transactions within minutes via MALI crypto technology. This will enable us to:

Receive revenue directly from prequalified borrowers who agree to Upstart’s loan arrangements

Receive revenue from our partners (e.g. Redfin, Mercedes)

Other Benefits:

Easier penetration into high-value markets such as Mortgage, Auto loan & lease and Personal loan:

⁠

Partnerships will enable more brand exposure and connect drastically improve Customer Acquisition in each of the high-value markets.

Channels

Partnerships:

Real Estate companies that deal directly with home buyers

Automobile companies and dealerships

Banks

Social Media Channels and SEO:

Optimize popular channels such as Instagram, FB, Google etc. advertise and direct potential borrowers to Breezi landing page

⁠

Open Issues & Questions

⁠

Add a topic

⁠

Questions & Unknowns

Questions & Unknowns

Questions

Author

Upvote

Downvote

Is the partnership route opportunistic to them enough to purchase our product?

Do we already have an ETL mechansim to make sure data is normalized when consuming additional sources

From a legal persepctive, do we have to consider/abide by any new rules

What types of loans will we offer and what will we not?

Are we resourced enough Across the AI/ML, Engineering, and UI/UX team?

We don’t really know how much users are really willing to pay for his service

How much training and hiring will we need to do for the approval team?

In order to verify identity faster, can we leverage the AI/ML platform to scan a valid ID to speed up the process

There are no rows in this table

⁠

Resources

⁠

Average U.S. Consumer Debt Reaches New Record in 2020⁠

⁠

What Are Basis Points and How Do They Affect Your Mortgage?⁠

⁠

Want to print your doc?
This is not the way.

Try clicking the ⋯ next to your doc name or using a keyboard shortcut (

CtrlP

) instead.