Microfinance: profit that has not happened yet

“Fourth year of growth in a row. Lending doubled, the report showed 1 493 824 ₴ net, I was already planning a third office. And then my partner asked: Inna, where is that money? And I realised part of it does not exist yet.”

Inna Shevchuk co-owns a microfinance company in Bila Tserkva. Short loans against no collateral at all: 12 000 ₴ for six months, 4.5% a month, a decision in twenty minutes. Eight years in the industry, four of them in her own business. Nine people, one office, growth every year.

What a single loan looks like

The client takes 12 000 ₴ for six months. He pays 4.5% a month, which is 3 240 ₴ of interest over the whole term.

Annualised, that is 54%. The figure looks large, and it is exactly why microfinance is thought of as an easy business.

The decision is made in twenty minutes, and that is part of the product. A client comes when he needs the money today, not next week, and it is that speed he is paying the rate for — a rate that would look wild in a bank. A bank decides nothing in twenty minutes.

The rate is not the margin

Out of those 3 240 ₴ you have to take off the cost of acquiring the client, the work on the problem cases, the fixed costs of the office and, above all, the part of the principal that will never come back. The rate shows the price of money for the client. It says nothing about how much is left for the company.

• principal: 12 000 ₴
• term: 6 months
• rate: 4.5% a month, that is 54% a year
• interest over the full term: 3 240 ₴

A year in lending

Over the past year Inna issued 2 400 loans. That is 28 800 000 ₴ of principal through the till.

Lending ran evenly, two hundred a month. This product has no seasonal dips at all: people need money the same way all year round, and that is one of the reasons Inna thought of the business as predictable.

Interest accrued: 7 776 000 ₴.

Interest received: 7 205 760 ₴.

Those two figures live in different places in the system. The accruals are done by the accounting software from the payment schedule; the receipts come from the till and the bank. Nobody ever asked for them to be reconciled monthly, because the difference looked like a technicality.

Accrued and received are different figures

The difference of 570 240 ₴ is the interest of the people who stopped paying halfway through the term. It accrued in the system, went into the monthly reports as income and never arrived in the account. That is 7.3% of everything accrued.

There is no bookkeeping error here. An accrual is an accrual, and it is done correctly.

What one client costs

Acquiring a single loan costs Inna 480 ₴. That is advertising in the local groups, the sign over the door, part of the salary of the people at the front desk, and the calls to those who came in and did not borrow.

Over a year, across 2 400 loans, that is 1 152 000 ₴.

The cost falls equally on every loan issued, because at the front desk you cannot tell which of them will turn out to be the 11%. Screening them out in advance does not work: that is precisely what scoring is for, and precisely why it costs what it costs. Until then, every client at the till looks equally good.

• acquiring one loan: 480 ₴
• over a year across 2 400 loans: 1 152 000 ₴
• interest from one loan: 3 240 ₴
• acquisition takes 14.8% of that interest

Who does not pay back

Inna's default rate is steady: 11% of loans never return the principal.

Out of 2 400 issued that is 264 loans. The good ones, accordingly, are 2 136.

The figure is steady for a reason. It is the share a specific product produces on a specific audience: a short unsecured loan in a town of a hundred thousand people. Change the product and it changes too, but until then you can plan on exactly that number.

“Eleven percent is my number, it has not moved for three years. I am even proud of it, because the neighbours do worse. The problem is not what it is. The problem is when it arrives.”

Someone who never repays the principal usually pays for the first two months. So the company does get 1 080 ₴ of interest out of him, and then he disappears.

Two months of payments is not chance, it is behaviour. A person takes the loan meaning to repay it, makes the first instalments, and then the very thing that sent him looking for money in the first place happens: the income stops. That is why default has such a steady profile in time.

What is left of the principal

264 loans at 12 000 ₴ each is 3 168 000 ₴ of lost principal.

Recovery works and brings back roughly a third: 1 108 800 ₴. The collectors' work takes 22% of what they bring in, that is 243 936 ₴.

Two million fifty-nine thousand two hundred hryvnias that simply did not come back. This is not a cost you can cut, and not a line you can renegotiate.

What recovery actually does

Inna's collectors are her own, not outsourced. They work for 22% of what they bring in, and they return about a third of the lost principal.

Recovery does not bring the business back to zero

Out of 3 168 000 ₴ of lost principal, 1 108 800 ₴ comes back, and 243 936 ₴ of that goes to the collectors themselves. So recovery closes a little over a quarter of the hole, not half of it and not all of it. The remaining 2 059 200 ₴ the company simply loses.

“My own collectors work out cheaper than outsourcing and they work more carefully. But I know exactly where I stand: they bring back a third. Anybody promising more is either lying or using methods that bring an inspection along afterwards.”

The year in figures

The report follows every rule. The interest was received, the costs are actual, the write-offs are backed by decisions. Not one line here is an estimate or an assumption, and that is exactly why Inna never once questioned it.

Sixteen percent of the interest. Out of every hryvnia that came in from clients, sixteen kopecks stay in the business.

The loss of principal is 1.79 times larger than the whole annual profit.

The reserve that was never made

Now let us put the question differently. How much of every loan should have been set aside on the day it was issued?

The answer is simple: 858 ₴. That is 11% of the principal, multiplied by the share that cannot be recovered.

Banks do this as a matter of obligation, and not out of their own goodness: the regulator makes them reserve against every loan before anything has gone wrong. A microfinance company carries no such obligation. So the reserve either exists in the owner's head, or it does not exist at all.

A reserve is not caution, it is arithmetic

858 ₴ per loan is 7.15% of the principal and 26.5% of all the interest the client will pay over six months. A quarter of the rate does not belong to the company from the very start: it belongs to the 11% who will not come back.

Over a year that is 2 400 × 858, which is exactly 2 059 200 ₴ — the same money as in the table above.

“When I saw 858 ₴, my first thought was that it is a quarter of my rate. For four years I had thought I was charging 4.5% a month. In fact I charge around three, and the rest is insurance I never wrote out for myself.”

Why the report showed more

Everything above is the economics of one year of lending. And now the main part.

For the first two months the client pays, and that creates a false sense of things. In the report he looks like a normal borrower, in the schedule he is active, at the till he is a source of receipts. His 1 080 ₴ of interest has already come in, and there is nothing bad to say about it.

Default does not arrive together with the loan. It arrives in the fourth month on average, when the client stops paying and it becomes clear that he is not coming back.

Write-offs lag four months behind lending

A loan issued in the first month of the year defaults in the fifth — inside that same year. A loan issued in the tenth defaults in the next one. If lending is the same year after year, this changes nothing: as many write-offs move forward as arrive from the year before.

That is exactly what breaks the picture. When the pace of lending grows, the window of the current year catches the defaults of a smaller volume of lending, while the defaults of the new volume have not happened yet. The difference settles into the report as profit.

Inna doubled her lending.

Vintage against the calendar

Last year she issued 100 loans a month. This year, 200.

So what was written off this year were the defaults of loans issued over the last four months of last year and the first eight of this one. That is 400 loans at the old pace and 1 600 at the new one.

The arithmetic here is mechanical. The write-offs of this year are 11% of the two thousand loans issued in the window from last September to this August. The defaults that belong to this year's lending are 11% of 2 400, and they will land over the next twelve months.

Forty-four loans have not fallen yet. They will fall next year, and next year will then carry 343 200 ₴ of a loss that belongs to this one.

Why this is not a fiddle

Not one figure in the report was wrong.

What was written off was written off. What was accrued was accrued. The bookkeeping was done by the rules, and an audit would confirm every line.

That is where the trap is. A business standing still sees the truth in its report automatically: write-offs and lending are in balance. A growing business sees an overstated figure every year, and the faster it grows, the bigger the overstatement.

Growth on its own overstates profit

The faster the portfolio grows, the larger the share of its defaults that has not happened yet. Calendar profit always looks better than the truth, and that is not the consequence of anybody's actions but a property of the construction itself.

Doubling the lending overstated net profit by 29.8%.

What had been tried before

“My first impulse was to tighten the scoring. If the problem is in those eleven percent, then do not let them in. I even set aside a budget and bought extra data sources. And then I did the sums and stopped.”

The impulse was perfectly logical. If the biggest cost in the business is the people who do not pay the money back, then the most obvious answer is not to give them any. The trouble is that here the obvious answer does not add up the way it seems to.

Tighter scoring costs 420 000 ₴ a year: access to the databases, the checks, a person to read them. And it does work: default falls from 11% to 8.5%.

But it does not only cut out the bad ones.

The scoring that did not pay for itself

By Inna's calculations, scoring would have cut 200 applications out of 2 400. That leaves 2 200 loans issued and 187 defaults instead of 264.

• there would be 77 fewer defaults
• but 200 fewer loans
• interest received 6 724 080 ₴ instead of 7 205 760 ₴
• net loss of principal 1 458 600 ₴ instead of 2 059 200 ₴

The reason is in the structure. A good loan brings in 3 240 ₴ of interest, while a bad one takes away 7 800 ₴ of principal after recovery and gives back 1 080 ₴ in interest. Turning down one bad loan pays off, but not if you turn down two good ones along with it.

Net profit with scoring is 1 016 692 ₴. Without it, 1 150 624 ₴.

So the scoring took away 133 932 ₴.

The threshold where scoring works

One loan brings in 1 834.86 ₴

Work out the contribution of a single loan at a default rate of 8.5% — interest less the loss of principal, less recovery, less acquisition — and you get 1 834.86 ₴. That is the amount every needless refusal the scoring issues costs you.

To cover its own 420 000 ₴ and come out no worse than the current result, the scoring needs at least 2 273 loans issued. So it can refuse 127 applications at most.

And that is on optimistic assumptions. The calculation assumes the scoring cuts out exactly the people it should, and that the rejected applications do not walk to the competitor next door and come back later with a worse history.

Inna's figure was 200.

“The gap between a hundred and twenty-seven and two hundred looks like small change, until you see that one side of that gap is minus a hundred and thirty-three thousand. I recalculate that model every quarter now, because default and conversion move independently of each other.”

There is one more reason not to tighten the scoring, and the figures only show part of it. A rejected application is not a neutral event: the person tells his friends he was turned down, and in a town of a hundred thousand that matters.

What Inna changed

She did not tighten the scoring. Instead she did four things, none of which cost money.

• the reserve of 858 ₴ goes onto the loan on the day it is issued, not when the default arrives
• defaults are counted by the month of issue, not by the month of write-off
• the report shows both figures side by side: the calendar one and the vintage one
• the scoring model is recalculated every quarter together with conversion

The third office appeared a year later than Inna had planned. But it opened against a figure of 1 150 624 ₴ rather than 1 493 824 ₴, and the money for it really was there.

Not one of these changes needed new software or extra people. All the figures were in the system already — nobody had simply put them into the right month.

The money does not disappear — you simply do not see it

“Eight years in this business, and I still think the most dangerous month is not the one when everything goes badly. The most dangerous one is when lending has doubled, the report shows a record, and you believe it, because it was calculated correctly.”

Nothing changed in Inna's business: not the principal, not the term, not the rate. Default is still 11%, recovery still brings back a third, lending is still 2 400 loans. The only thing that changed is that the loss now sits in the same month as the lending that produced it.

How to see this in your own portfolio

Everything above rests on one cut the monthly report does not make: a loss belongs to the month the loan was issued, not to the month it was written off. While write-offs follow the calendar, every increase in lending automatically overstates profit — and the faster the portfolio grows, the larger the earnings look that have not happened yet. When issuance, payments and write-offs line up by vintage on their own, the 858 ₴ reserve per loan becomes a line rather than a surprise — which is exactly why owners of small and medium businesses use Finmap: to see the financial picture instead of reconstructing it after the fact.

Read also

• Pawnshop: 44.7% of capital sits on the shelf
• How to Get Clients to Pay on Time: 6 Ways to Clear Overdue Invoices

Topic foundation: Management Accounting Explained: What It Is and Why Owners Need It