Assets and Liabilities: A Tool for Wealth Management
The Assets and Liabilities products use transaction lines to understand end-users' investments and personal assets, and the returns on those investments.
At Okra, we are always cooking, trying out different recipes and formulas to create amazing products and services that businesses and end-users will find helpful. We are proud to announce Assets and Liabilities!
The Assets product uses transaction lines to understand end-users investments and personal assets, and the returns on those investments. It uses a collection of data analysis and machine learning modes to understand when the initial investments were made, the amount, and the subsequent ROIs, dividends, and appreciations on the investments.
Liabilities (Loans) product models end-users financial obligations such as rent, loans, and mortgages. It also shows how frequently a user meets these obligations. A key element of the liabilities product is loans which present all loans previously taken by the user, their repayment schedules, and any active ones yet to be repaid.
The Assets and Liabilities products can help assess the wealth and risk profile of a user.
Use Cases:
Assets and Liabilities products are helpful for wealth management and lending companies. However, any company/institution could benefit from information about their users' assets/investments, loan history, or active loans. Possible applications include asset lookup, asset transfer, and wealth management: debt purchase, debt history lookup, and risk profiling.
Product Details:
At the core of Assets and Liabilities, and many other Okra's products is the Okra Natural Language Understanding (NLU) engine - a group of robust text processing systems capable of understanding the meaning of words and recognizing special keywords/named entities. The Okra NLU engine is the base pre-processor for many transaction text analyses at Okra.
At the Okra NLU level, a group of machine learning models understands the meaning of transaction texts and identifies keywords/named entities. Internally, Okra NLU maintains Company Predictor, a machine learning classifier that can recognize company names. Using the list of companies in Nigeria and Company Predictor, NLU is able to assign the recognized company names as the sources of the dividends, stock (asset) liquidations, or liabilities.
A hypothetical analysed batch of transactions returned from Okra NLU can be seen below:
A typical sample from the Okra NLU can contain the type of transaction, the sub-type and the beneficiary of the transaction. It also contains the source bank, beneficiary bank, and the source of the transaction. An interesting keyword in the NLU engine response is chunks which identifies transactions that belong together in a larger investment disbursement, investment return, loan disbursement, or loan repayment group. Chunks contain the internal IDs of other transactions that belong together and will help the machine learning models (explained later) to group these disbursement chunks together.
Modeling Assets & Liabilities
The NLU batch response is sent to the Assets and Liabilities machine learning models, which understand the intricacies of the pre-processed transactions and correctly classify them as either "assets", "liabilities" or "others" with stunning accuracies.
Precision, Recall, F-score, and accuracy are used to evaluate the performance of the machine learning models. The summary of the performance is shown below:
Integrating to your product:
To integrate Okra to your app or website, read our extensive documentation to see relevant endpoints, code snippets, and more. If you are already familiar with setting up and using Okra, then head to Assets & Liabilities sub docs to understand the request and response objects.
Request Objects:
To use the Assets and Liabilities products, you make a POST request to the URL passing customer ID string. You will also have to pass the secret key attached to your Okra account.
Response Objects:
Asset Response:
A typical response from the Investment product can be seen below:
The first key in the response is the confidence score which measures the performance of the model's ability to properly identify and classify asset-based transactions.
Asset (Investment) Groups
The asset response is broadly classified into two - dividends and investments. Dividends come mainly from publicly traded companies identified by the Okra NLU with data ingested from the Nigerian Stock Exchange (NSE). This particular user has no dividends. However, he has a few investments in the Meristem Money Markets Fund.
The investments group is further split into deposits and returns. Deposits are chunks of debit transactions that occurred when the investment was made - as understood by the Okra NLU. Conversely, returns are chunks of credit transactions that are essentially the benefits of the investments. Each deposit has a transaction date, a recipient of the deposit, and the amount deposited. Each return also has a transaction date, the source of the return, and the amount credited. There are also no_deposits, which is the number of times deposits were made (length of deposit chunks), no_returns, which the number of time returns was credited (length of return chunks), total_deposits, which is the total sum of the investments, and total_returns_received, which is the total sum of returns received.
Matched Investments
An interesting key in the investment return object is matched investments. Matched Investment group's investments and returns by company names making it possible to access the total deposit and return sums in one place using only the company names. This is helpful for an overall summary of investments and returns grouped by company names.
Liabilities (Loans) Response Object
The liabilities response object is grouped into three major categories - mortgage, rent, and retail loans. Mortgage handles mortgage payments, rent identifies the yearly rent obligation, and retail_loans gives a detailed breakdown of the user's loan history.
Matched Loans
Like investments, liabilities also have "matched loans" which are loans grouped by company names. Each matched loan has a "repaid" key that indicates whether the loan has been fully repaid or not. This is particularly useful to lending companies who can use this key to access all the active loans yet to be liquidated by this user.
This particular user owes Okash a total of N2,500,000 and has only paid back N137,000.
More Data Science Products:
While you are at Okra docs, feel free to check out some of our other amazing products like Income and Spending Patterns. Check out the full product suite and find what you need here. And if you'd love to see a demo, we have you covered; book your exclusive session here. Let's build the unimaginable together!
