Mobi-Scor

Many credit-worthiness & insurance risk methodologies consider traditional payment performance data as a primary input into their risk models. Payment performance data includes but is not limited to credit card, home loan, and vehicle finance repayments. 

By definition, any business or individual who does not have a formal credit history is excluded from the financial services ecosystem. This further skews credit-worthiness models making them more biased and discriminatory against minorities.

Our approach is to use behavioural metadata gathered from mobile devices and the web to predict risks across the credit lifecycle of the consumer or business. This includes credit risk, fraud and data augmentation use cases. 

By not assessing personally identifiable data we are generally compliant with local and international data processing and storage regulations.

We believe our methodology & technology could not only provide credit-worthiness scores for those who don't have one, but improve and augment existing credit scoring technologies irrespective of whether the individual or business is a fat, thin or no-file customer providing richer data to reduce bias.

Many credit-risk models consider an applicant's ability to pay. However, they don't consider their willingness to pay. 

By adding whether someone is willing to pay many credit and underwriting models could be enhanced to score a greater number of businesses and consumers increasing financial inclusion by using quantitative non-subjective data.

South Africa has a population of 60 million people with 65 million smartphone subscriptions. 17 million people have no credit scores whilst 27 million existing credit consumers could have their credit scores enhanced with alternative behavioural data.

As limited or no alternative data is added to credit decisioning models this contributes toward 50% of formal loan applications being declined.

Although the informal MSMEs market is worth over $5.4 billion many of these businesses cannot access affordable and unbiased financial services which impact access to essential business funding. This is a result of high levels of fraud, a lack of identity verification and a lack of traditional credit payment profiles.

314 million individuals residing in Sub-Saharan-Africa (SSA), Middle-east & North Africa (MENA) are underbanked or have no credit score. This is despite the fact that SSA has an expected smartphone penetration of 61% by 2025 while MENA has an 84% smartphone penetration rate.

Globally 1.8 billion people are underserved or excluded from affordable, fair and non-biased financial services.

The analysis of behavioural mobile device & web metadata allows for anyone to acquire a risk score anywhere in the world at any time. 

Since we collect first-party data there are fewer limitations compared to open-banking data availability, fragmented telecommunication data coverage and geographically reliant data aggregators.

Metadata collected is the same across any mobile or web device and a score can be generated regardless of the economic climate.

Our mobile device and web metadata scoring product is called Mobi-Scor. 

Mobi-Scor can be embedded into any existing financial services provider's process by embedding a low-code mobile app or web SDK.

Both mobile and web SDKs do not require the financial institution to change their existing processes or application layouts.

When an individual begins a credit application for instance on a web page every interaction is recorded. This includes swipes, keystrokes, copy paste actions among many others. 

If an individual is applying via a mobile application, they would grant permissions to storage, and device data before commencing with the application.

Mobile device metadata is then extracted from the mobile device in a similar fashion as the web application. Metadata would include contacts, calendars, and storage information among several other variables.

As soon as an individual commences an application a unique application ID would be sent to the financial services provider. 

Mobi-Scor would not record any personally identifiable information. All metadata is binary by nature ensuring the applicant's identity is private at all times. 

Mobile device and web metadata is uniform across any applicant, geography or language anywhere in the world. This allows for maximum coverage and risk profiling.

Once the metadata is collected it is run through a machine learning engine which provides a predictive score for the applicant.

The machine learning model can be calibrated for several many events across credit risk, fraud and enhanced customer profiling use cases.

Here are some examples of how a machine-learning model could be calibrated depending on the use case:

Credit risk model:

- Predict first payment default

- Probability of default

- Predict willingness to pay

Fraud model:

- Predict persistency rates (whether a policyholder will pay their premiums regularly)

- Claims fraud

- Policyholder KYC 

Enhanced customer profiling:

- Predict product uptake rates between different products  

   https://youtu.be/AggqUY7w0bQ           

There are two beneficiaries of our technology. 

B2B: Providing financial services providers with new data that are predictive of credit risk and fraud while enabling them to enrich their existing customer profiles 

Benefits for B2Bs are:

- Enabling financial services providers to expand into new markets by providing predictive data on consumers and businesses they otherwise would not have access to. 

- Providing credit scores for new to-credit businesses and consumers for short-term lending.

- Providing credit & fraud risk data to informal economies.

- Enabling micro-insurance providers with data to create appropriate insurance products for specific use cases (agriculture, asset finance, stock advances).

- Provision of Data-as-a-Service for BNPL (Buy Now Pay Later) providers, Neobanks, Challenger Banks, E-hailing services, cross-border providers, mobile wallets, airtime credit & micro-loans.

Benefits for B2C customers who are traditionally unable to get credit, insurance or micro-enterprise business finance.

- Individuals and micro-businesses who don’t have credit scores and are underbanked. 

- Minority groups

- Foreign nationals & refugees 

- African Americans 

- New to credit customers

- Woman-run households

In South Africa, there are 44 million people who could benefit from Mobi-Scor.
In Sub-Saharan Africa, there are roughly 1.1 billion people, 189 Million who're under-banked or have no credit score.

In MENA (Middle East & Northern Africa) there are 486 million people, 125 million of which don't have credit scores. 

In America, there are more than 45 million consumers with limited or no credit profiles. This is a conservative number of consumers.

As the CEO of Vizibiliti I left home when I was 17. I was homeless for a period of my life and when I tried to apply for any form of credit even though I had a formal job I did not have a credit profile which excluded me in my personal capacity from accessing any financial service.

I started my first company when I was 19. In my business capacity I was further excluded from the financial ecosystem as my business had not been trading for 3 years, I was young and did not have a university degree since I couldn't afford to go to university. This resulted in me being considered as a minority group.
For 14 years I have had direct B2C experience with the challenges faced from the perspective of a small business owner and an individual consumer. 

During the past 7 years, I've worked with multiple B2B financial services institutions across multiple credit verticals including vehicle and asset finance, mass market insurance products, traditional banks, and one of the largest microlenders in South Africa that's provided micro-loans to more than 800,000 South Africans. The majority of these customers are low-income customers who cannot get credit from traditional financial institutions.

Mark Young our CRO (Chief risk officer) has extensive B2B and B2C credit experience.

He was the CEO of Old Mutual Financial Services which are one of the largest personal lending institutions in South Africa. Old Mutual also had the largest base of mass-market insurance customers in the country. 

He was the CEO of Bayport which provide personal loans across Africa, North & South America.

He was also the CEO of Get Bucks which provided loans across 7 African countries. 

Mark's knowledge of credit granting, decision processes and risk-scoring including the impact of alternative data give him a comprehensive view of the challenges for consumers and lenders alike.

Our internal AI team are experts in the development of predictive alternative credit granting and insurance scorecards. We've successfully developed an alternative credit-granting scorecard for one of the biggest microlenders in South Africa & for a large debt-collection business.

Theuno de Bruin is the Chief Technology Officer. He is responsible for the development of both front and back-end software.

He has extensive experience in developing bank-grade software. He worked at First National Bank which is the second largest bank in South Africa.

He further managed teams and built software for Edcon which is one of the largest retailers in South Africa.

He has a true understanding of the financial services industry and how to effectively develop world-class software solutions.

We're processing 50,000 low-income credit applications and roughly 250,000 debt collection profiles a month through our alternative risk scorecard.

Networks & partners who we could integrate our solutions with are critical to our growth. We have goals to expand into Sub-Saharan Africa and Middle East & North America in the next 12-18 months.

We want to engage with partners and potential customers who are serious about innovation and are willing to try new approaches when solving complex problems.

We further want to develop our partner network which already has a presence in Africa, the USA and the Middle East.

From experience having partners who understand the local culture and economic climate can add essential insights when scaling into those markets.

We don't believe we know everything and can do everything alone. We're always interested to engage with those who have experience where we do not. We have found peer groups are a great way to learn from others who've solved for problems like product-market fit, growth and product development.

We have found that many organisations claim to support social initiatives where they are really only focused on financial gain. The Solve community clearly has alignment between running sustainable businesses which can make a tangible impact on communities and locations they operate in.

Many financial institutions are generally constrained by three key inputs into their risk analysis processes.

One of the significant problems that financial institutions face is the type, quality and availability of the data which is used as inputs for credit and fraud modelling.

By example:

Bank statement data via open banking - Key problems include whether there are agreements between banks, aggregators and API hubs. This is further limited by the number of banks that offer API's in their environments. 

Another limitation is whether the business or consumer has a bank account and whether they use that bank account. Therefore data coverage is limited. 

Utilities and rental data - Coverage is limited and is reliant on whether a utility provider reports payment data. In Africa utility and rental data is disparate and generally has poor data coverage. 

South Africa is a good example. Rental & utility data within informal settlements & rural communities is generally paid in cash and isn't recorded. 

There are only 1.4 million traditional rental payment profiles available out of 60 million people.

Any form of credit bureaux/central repository is reliant on whether 3rd parties submit updated consumer and business profiles.

By using mobile device and web metadata one is able to assess 100% of consumers and businesses. The data is first-party in nature, privacy compliant and doesn't require 3rd party data sources 

Unlike the above, there are far more reliable and standardised data points one can gather from mobile devices and web metadata which exceed any of the traditional/alternative data sources.

There are 65 million smartphone subscriptions in South Africa. There are many more mobile devices than alternative data sources.

The second is using new methods of risk modelling specifically AI and machine learning. The third challenge is the explainability as to why a specific decision was made.

Many financial institutions use scorecards which are still based on logistic regression models. These are outdated and don't provide machine-learning capabilities or insights.

Our machine-learning models can provide clear reasons why an application was accepted or declined.

This provides the financial institution with advanced scorecard capabilities with the ability to include quantitative mobile device & web metadata.

Another factor to consider is data privacy. Our approach of collecting binary metadata without recording any personally identifiable data means our process is generally compliant with data privacy laws.

Our scorecards can be calibrated for many use cases across finance, insurance, investing, saving among many others.

Alternative mobile device and web metadata can also be used for credit risk, fraud and customer profile enrichment.

This year:

- Develop and deploy a new alternative credit scoring scorecard which combines all traditional data with mobile device and web metadata to create a new type of risk score.

- Run an A/B test with one of the biggest microlenders in the country to identify the specific mobile device and web features which are most predictive of defaults.

- Develop an embedded environment which is low-code and can embed a mobile app and web SDK into most digital & traditional financial services institutions.

- Develop and deploy the SDK

- Deploy the SDK within the South African insurance industry.

- Secure partnerships with large organisations that have a high number of applications from both fat, thin and no-file applicants.

2024

- Prepare our organisation for expansion outside the South African market. That will likely include new roles, hiring processes, executive appointments and the implementation of scalable systems.

- Begin expansion into new markets (Sub-Saharan Africa and Middle-East and North Africa). 

- Increase the number of customers who could utilise the embedded finance solution. This would include BNPL (Buy-Now-Pay-Later), Challenger banks, Neobanks, E-hailing, Sharing economy businesses, and microinsurers.

- Increase the speed at which financial institutions can activate, deploy and acquire results from our new scorecard, mobile device and web SDK.

2025 onward

- Continue investment into R&D and product development to identify additional sources of scalable metadata which are complimentary to our solution.

- Further embed our technology into new industries including investments, savings and wealth generation.

- Explore expansion into LATAM countries.

Number 5 Gender Equality 

Provide risk scores to many more individuals and businesses in need of financial services especially youth and women irrespective of geography, creed, colour and language.

Metadata will further support a non-biased view of the applicant.

Number 8 Decent work and Economic Growth

Provide predictive risk models for thin and no-file applicants with quantitative data points from anyone geographically especially those in semi-rural and rural areas with no formal trading histories or bank accounts.

Number 9 Industry, Innovation, and Infrastructure.

Provide consumer and business applicants irrespective of high, medium or low-technology industries the ability to acquire a risk score for lending, insurance or other financial services.

Number 10 Reduced inequalities 

Provide financial service providers with an unbiased score of any applicant seeking a product or service. 

Since the alternative risk scorecard isn't recording male, female, gender etc. scores are generated exclusively on behaviour and not personally identifiable data.

Provide consumers and businesses with an unbiased risk score. 

This includes measuring the willingness of an individual or business 

Number 17 Partnerships for the Goals

Develop innovative embedded finance products e.g. credit finance based on transactional money transfers.

Provide additional quantitative behavioural metrics to strengthen risk models enabling financial institutions with the ability to more accurately determine creditworthiness based on affordability and willingness to maintain a financial commitment.

Assist a growing number of individuals and businesses who have access to mobile and web devices the ability to engage with financial services.

By providing businesses and individuals with behavioural scores derived from mobile device metadata we provide financial service providers the ability to score 100% of their applications. 

Alternative risk-scoring scorecards can consume this metadata and provide a richer view of thin or no-file credit profiles.

Short-term measurement 

- Identify which mobile device and web metadata features are most predictive of positive and negative risk. 

- Combine those features with internal behavioural payment profiles, application data and credit bureaux information. This will create a new type of predictive scorecard which considers both the ability to pay and the willingness to pay.

- Monitor a loan/insurance payment behaviour over a period of time. By example, a short-term loan such as a 1-month loan would only take 1 month to determine the characteristics of a high or low-risk consumer or business application using mobile device or web metadata.

Medium-term measurement

- Allow financial services providers access to a low-code embedded SDK which can easily be incorporated into their current operations.

- Enable the financial services provider to score new thin or no-file applicants which will increase the volume of applications and provide additional willingness to pay variables into their existing scorecards.

- Monitor loan/insurance payment behaviour to determine features and characteristics of 1) Successful vs. unsuccessful applicants 2) Positive and negative payment profiles 3) Monitoring the impact on debt collections both insourced and outsourced.

Long-term measurement

- Create an aggregated dataset of all mobile device and web metadata collected for each industry and provide a predictive score for a specific industry. 

- Enable financial services institutions to create and manage hosted risk scorecards within our environment.

- Provide an Infrastructure-as-a-Service platform to all financial service providers at scale. The platform can connect to multiple sources of anonymised data sources via one managed API hub.

The current alternative credit machine learning scorecard  developed for one of the biggest microlenders in South Africa includes data from:

- The applicant filling in their application form

- Credit bureaux checks

- Affordability assessments

- Internal payment performance data

Our alternative machine learning scorecard increases loan approvals from a normal credit bureaux model which approves 22,900 loans to 28,100 loans. Roughly a 26.1% increase in loan approval with a similar default rate of 15%.

We expect these results to improve and augment the existing model with the addition of mobile device and web metadata.

Our alternative scorecard solution within the debt collection industry has a 90%  accuracy rate of predicting RPC (Right Party Contact) which is essential to contact the right debtors at the right time. 

This encourages debtors to pay off their debt as soon as possible enabling them to receive a discount on their remaining debts.

The process of submitting the data to our machine learning models is simple and efficient with results returned in a short period of time.

All our solutions are cloud-based enabling us to scale with any increase in application volume.

Additional unique advantages of our process are:

- Identifying loan applications which are currently declined which should be approved by the existing scorecard.

- Identifying current repeat applications which should be declined to avoid reckless lending.

- Ensuring that we don't use personally identifiable data within our data collection and processing process.

- Ensuring all applications comply with affordability standards defined by the regulator.

Our mobile device and web metadata platform provides the following benefits:

- 100% of mobile device and web applications could be scored.

- No additional information e.g. psychographic data is asked from the applicant while submitting their application.

- Since the data collected is metadata the technology is can be deployed globally irrespective of language or geography.

- As no personally identifiable data is collected the input data is non-bias

- Since no personally identifiable data is collected the technology meets data privacy standards.

The primary factor that will always be more important than any technology or product will be our people.

Our team is diverse with each member having a different skill set, upbringing, and family environment.

We have members who are from different religious backgrounds including minority groups.

Our CEO has been homeless before, excluded from the financial ecosystem and has first-hand experience in how current risk-scoring is performed and how the company could make a measurable difference within financial inclusion.

The CRO has extensive experience in insurance, credit analytics, scorecard modelling and behavioural economics. He's been the CEO of one of the largest microlenders in South Africa which has a global footprint in Africa, Columbia and Mexico.

We are developing solutions which are inclusive, globally relevant and scalable. Our goal is to improve the entire financial services environment especially for those who don't have fair and equitable access to it.

We are a B2B SaaS business.

One of the primary reasons for this choice was to ensure that individuals and businesses applying for a financial product do not bear any cost.

Businesses paying for our services benefit from our technology since they can risk-score many more applications with a lower risk of a negative outcome (e.g. default).

As our mobile device and web application solution grow they will have further capability to score underbanked, thin and no-file consumers or businesses.

Financial services providers will also be able to personalise pricing making the pricing of their products more fair and accessible while remaining compliant with governance regulations.

Financial services providers pay for our services on a monthly basis with a fixed fee during a 12-month contract

We generate revenue by charging financial services providers on a monthly basis over a fixed term.

This enables the financial services provider to know what their costs will be over a long period of time.

Alternative data solutions are billed on the number of API calls. This is calculated by assessing the average number of monthly applications over an extended period.

Should the number of monthly applications exceed that of their pricing tier they then pay an out-of-bundle pro-rata rate.

Consumers and businesses applying for financial products will not have to pay any additional fees for their application to be assessed.

Our key cost centres are human resources and product development.

Sales and marketing expenses for B2B sales isn't really high therefore our CAC costs are low.

We have been self-funded thus far. We rely on existing customers to fund our operation. For us to achieve & accelerate our software development and growth goals we need external funding.

We intend to use the following strategies to acquire more capital:

- Access grant funds that align with our social and financial objectives.

- Raise capital from investors where we can add value to their existing portfolio companies.

- Upsell products to existing customers.

- Acquire new B2B customers.

- Align with partners who can sell our services into their existing customer databases. A commercial model accompanies each agreement.

If we consider our debt collection client we initially built the model which took one month to build. 

The only costs we incur are human resources, minor software development and cloud fees.

Once the model is built and the data pipelines are structured we only incur cloud costs which are a fraction of what we charge a financial services business.

A similar financial model applies to the mobile device and web metadata solution. 

There's a technical framework which is generic across different operating systems and devices. 

The same architecture can be embedded across many different applications from multiple financial institutions.

Once the bulk of the internal development has been completed the internal costs would be to maintain existing integrations and keep the technical stack up-to-date with changes from app stores.

We have previously generated R1,600,000 in grants and consulting.

We currently generate recurring revenue on a monthly basis.

We have won multiple prestigious awards including:

Forbes Africa-30 Under-30

Forbes Africa 30 Under 30 chosen from hundreds of businesses across the African continent.

Finalist in the Santam Insuretech Challenge 2019

Finalist Nedbank / Launch Lab Demo Day: 2019

Knife Capital* Venture Capital Grindstone Program 2018/9:

1 of 12 selected scale up’s out of 200+ submissions from around South Africa.

*Knife Capital is one of the only venture capital companies in South Africa to have successfully exited businesses to VISA for $100 dollars and Garmin for a similar amount.

Viva Tech Paris 2018:

Viva Tech Verizon Alternative Credit Scoring Award.

Nation Builder Social Innovation finalist 2018:

Alternative Credit Scoring solution for the Nation Build Social Innovation Award.

Nominated for the ABSA Innovator of the year 2018:

Nomination for the Alternative Credit Scoring award for excellence as an industry leader.

Awarded Mercedes-Benz Financial Services Alternative Credit Scoring Pitch 2018:

Nominated for the AABLA Innovator of the year award: 2018

Nominated for the Shark Tank award: 2018

Attacq Commercial Real Estate award for excellence in retail innovation: 2017

Launch Lab Commercial Real Estate overall winner 2017:

Mercedes-Benz Predictive Manufacturing Winner 2017:

Innovation excellence in predictive manufacturing for the Mercedes-Benz East London centre of excellence plant.

Distell Predictive Alcoholic Shelf-life finalist 2016:

Finalist in the Distell Alcoholic Shelf Life Award. Tshwane Cable Theft Award 2016:

Predictive cable theft platform