2. Loan Matching: Procedure and Methodology
When an amount of money X is loaned from a lender to a borrower for a period of time T, then a loan event has happened. A match was found between a borrower B* who agrees to pay an annual percentage r% for the benefit of having the loan, and a lender, L* who agrees to offer the loan, and be paid for it an amount commensurate with an annual interest of p%.
The gap (r-p) reflects the gain of the loan match maker. (the digital landing initiative).
We consider a situation where over a reference period of time, Tr, a loan match maker, LMM is asking for would be borrowers and would be lenders to inform it of their wishes.
A borrower will say: "I wish to borrow an amount B for a duration Tb, starting at date/time point tb1, and I am willing pay for it at an annual rate r%."
A lender will say: "I wish to lend an amount L for a duration Tl, starting at date/time point tb1, and I am setting my price for the loan to be p% at an annual rate."
At any given point of time t, there will be Nb(t) would be borrowers, and Nl(t) would be lenders. Each borrower and each lender came up with their own set of parameters for amount, duration, start time, and interest payment percentage (rate). The task before the aspiring loan match maker is to match lenders with borrowers so that few as possible borrowers will be left without a loan and few as possible lenders will be left without a borrower.
For simplicity, per the following analysis, we assume that the match maker is fixing the values of r, and p -- the paid and earned interest rates, to insure profitability for the matchmaker. So now all the participants comply.
We define a loan match variable, M, as the multiplication of an amount X that is being loaned /borrowed for time duration T:
We can set up a unit for loan matching, for example $1000 * 1 day will be one option, namely $1000 loaned for a day. Or finer levels. $100 * 1 hr, for very short very small loan events, or the opposite, much larger: $10,000 * 30 days, as is convenient.
Payment for the loan and earning from the loan, as well the loan matchmaker profit from the matched loan all are proportional to the match unit M = XT.
Similarly we define the borrowing load:
where the summation is over all the N
b borrowers. Namely: summations of the individual requested loan amounts. B
1, B
2, ,.. time the loan durations. T
b1, T
b2,...
Similarly we define the lending load, Q
l
where the summation is over all the N
l lenders. Namely: summations of the individual offered loan amounts. L
1, L
2, ,.. time the loan durations. T
l1, T
l2,...
Both summations are conducted over a reference frame, Tr, and include posted borrowing requests and posted lending offers that start and conclude inside the reference period.
Note that both the would be borrowers and the offering lenders submit their requests and statement as posting delivered to the attention of the DLI. There is a built in advantage for sending all the postings to the DLI and not making them public. If made public then individual borrowers and individual lenders will get together to a deal that may be unfair or suboptimized to one party or both, and it will deny the lenders the financial security offered one way or another by DLI.. Optimized loan build up from a large set of postings and optimized loan breakdown to a large set of counter postings is a complex mathematical task, best conducted by the DLI. The DLI profit depends on the amount of matched loans so it is in its interest to seek optimized matching from the available posting.
Since a loan match requires that every dollar offered for loan for every hour, is matched by a requested dollar to be borrowed for one hour, we can write the
basic matching formula:
However, the starting date for the various postings can be such that the maximum match per a given combined load from borrowers and lenders will be less that the 'best':
The basic matching formula suggests that any gap between the borrowing load and the lending load is a waste, the extra load will not be served. It is therefore that match making operation will apply the power of the market -- supply and demand -- to come as close as possible to equilibrium: Qb = Ql.
The power of supply and demand in cyberspace can be applied without any friction, ignoring all separating factors, like nationality and distance.
If we have Qb > Ql then one increases the value of %p and %r, so lending becomes more tempting and borrowing more prohibitive.
If we have Qb < Ql then one reduces the value of %r and %p so that borrowing becomes more attractive and lending becomes less attractive.
Since all are interested in minimizing waste, all parties seek an objective optimization between r% and p%. In the case where several DLI are in competition, they will compete and who offers the smallest gap between r% and p%.
We discuss below the basic procedures for match making and the AI empowered procedures to improve the results.
2.1. Basic Matching Procedure
The operation is driven by borrowers because, a borrower's request by default should be accepted, or rejected as is, while a lender posting can be partially taken.
Borrowers postings are examined per their start date, the earlier get attention early.
The matchmaker looks for lenders with posting fit into the borrower's start day, at any amount. The borrower's requested loan amount is built from available lenders who wish to lend a lesser amount, or from a lender with proper timing that wishes to lend a larger amount. All the lenders that assembled to service this early borrower are paying the money directly to the borrower. (in practice the money may go from the lender to the match maker (the virtual digital bank) from where it goes to the borrower -- in both cases using the digital cash instant transaction like in BitMint*LeVeL.)
When the entire requested loan is fit with proper lender's postings the first step is complete. The requested loan is digitally paid to the borrower. The borrower can use the BitMint digital money as is, or redeem it against legacy dollars, per their choosing. After the full amount of the load has been paid to the borrower, there are various time points where the matched lenders have finished their offered duration, and before that happens the match maker needs to find a proper lender to take over the lender that expects its money back before the loan is due. A given first lender L1 is finishing its lending mission at time point t1. When t1 comes to pass the match maker must be ready with another lender, L2 whose proposed loan is at least the amount serviced by L1. The replacement lender L2 pays the amount serviced by L1, X, but not to the borrower, rather to L1. This concludes the participation of L1 who made his lending and got his money back after the specified lending time. The earned interest will be paid later by the match maker.
Next the match maker watches for the next time point where any of the active lenders that is part of the group of lenders who support this borrower reaches its end of lending time, while the borrower's time demand is much longer. At that next time point the match maker will have to be ready with a new lender with proper terms. This new lender will pay the previous lender the amount the previous lender lent, and so again the retiring lender gets it money back at the time desired, being unaware that this lender is part of a loan match build up assembled by the match maker.
Continuity
Loan requests with long term durations will require lenders with long term lending offers. Alas, the edge of this DLI is that it captures short term lend offerings and builds them up to a long term loan request. The dilemma of the lender is that they face uncertainty regarding future surprises when cash and liquidity will be sorely needed. If they have their cash being lent out they are at a loss. That is why lenders are reluctant to tie up money for long periods, but prefer shorter periods with much less risk. Traditional banks find it hard to work with such short term lending. This DLI solution claims an advantage by opening up a whole new market based on short term lenders that engage their money through BitMint fast, irrevocable money transfer protocols. However these lenders typically can offer to lend now or soon, but are reluctant to commit for a lending at a point far into the future. The DLI then needs to close the gap between the most desired loans from borrowers -- long term loans -- and the most desired lending from lenders -- short term loans offered in the here and now.
To assure a borrower that their loan request for the requested loan time Tb will be satisfied, the DLI will have to rely on (i) a credible estimate of lending offerings to be posted long into the future, and (ii) a financial backup capacity to keep the operation going during periods where lending offering are too meager.
The DLI will need to promote the business to secure constant stream of lending offering. The DLI will have to come up with credible predictions as to the future state of lending offering. The DLI will want to rely on AI capability to predict future lending offering. Still, fluctuations will occur, and the DLI must be ready with a financial shock absorbing capacity -- ready to use cash, to serve as a lender of last resort to uphold a long term loan commitment the DLI made to a long time borrower.
In the beginning the DLI will need to be ready with a lot of lending capacity on its own. But as the word goes out and lenders are aware of the attractive options to let one's money work for one while one sleeps, more and more ready lenders will come forth and will make it less necessary for the DLI to pitch in its own money.
2.2. AI Empowered Matching Procedure
The complexity of matching hard to predict streams of postings from borrowers and lenders alike, makes the goal of optimal loan matching into a typical AI mission.
To prepare the data for AI, the DLI will keep record of the stream of postings over time. The overall time line will be divided to reference periods. R1, R2, ..... The postings that were submitted fully within a reference period will be analyzed "post mortem" to determine the best matching that could have been achieved. This Mbest value would be compared to the de facto M values achieved within each reference period. If they are similar then no need for AI boost. If not then the data for the growing number of reference periods will be given as raw material for AI supervised learning in order to find pattern with which to build a matching strategy with high matching score. This AI reference engine will become more effective the more reference periods there are to be analyzed. If the climate of the postings will change then the AI inference will track it by giving more weight to recent reference period relative to older ones.
If a market is serviced by more than one DLI then the various DLI will compete as to how well they use AI to increase the efficacy of the loan matching operation.
2.3. Elasticity
Given a situation where the B postings and the L postings are not in good match, it is advisable for the DLI to negotiate some changes in the terms. To check if a different sum will be Ok, to check if the requested or stated duration can be modified a bit, to check if the starting date can be different. And of course to re-negotiate interest rates. Some postings can come with a lot of elasticity as to such changes, and others not.
It is a role suitable for AI to propose a negotiation strategy. The DLI can offer better rate in exchange of terms elasticity.