A bank knows the cost of risk on its SME portfolio. It knows its net banking income per client, its cross-sell rate, its margin. It rarely knows what acquiring an SME client costs it — because that cost appears on no line.
It exists nonetheless, and it has three components, only one of which is measured.
The three components, two of which are counted nowhere
The cost of the completed file. The only visible one: relationship manager and analyst time on a file that reaches committee and passes. It is long — average processing time for an SME file runs into tens of days, often sixty to ninety — but it is charged to a real client, so it justifies itself.
The cost of the declined file. It consumes exactly the same resources, and it is charged to nobody. Yet African banks, asked why they do not lend more to small and medium-sized enterprises, cite first the lack of bankable projects at 40 % and insufficient collateral at 37 %. In other words: a significant share of the files opened never stood a chance, and they had to be appraised to find that out.
The cost of the file never completed. This is the heaviest and least visible. It is the documentary back-and-forth that stretches, the follow-up that goes unanswered, the file that falls asleep. There is no decision, no file and no client — only hours spent, and a relationship manager with less time for the next one.
In the CEMAC zone, four companies out of five operate outside official records, without certified accounts or auditable financial statements. That figure is the direct cause of the second and third components.
What an invisible cost produces
A cost that is not measured is not steered. It is worked around, in three perfectly rational ways.
Ticket size goes up. A 500-million file costs roughly as much to appraise as a 50-million one. With a fixed analysis budget, serving the large ticket is arithmetically superior. That is why the mid-market segment — the one every strategic plan discusses — is the first set aside, not by decision but by implicit calculation.
Retreat to the known. An existing client is appraised faster than a prospect, because account history replaces part of the file. Acquisition gives way to cross-selling, and market share freezes.
Unspoken rationing. The relationship manager learns to spot expensive files early and not to open them. It is individually efficient, and it collectively produces a selection whose criteria nobody ever set.
Why this becomes critical now
Two 2026 developments reduce the resource available to absorb this cost.
On 2 April 2026, the BEAC's Monetary Policy Committee suspended refinancing operations for medium-term credit intended for productive investment; the Governor confirmed on 29 June that the request came from the International Monetary Fund and sought the mechanism's progressive extinction. Less long resource means less investment credit, therefore a higher relative share of short operations — and the acquisition cost is then spread over smaller transactions.
COBAC regulation R-2025/02, in force since 1 January 2026, raises minimum share capital from 10 to 25 billion francs, with a first step at 14 billion on 31 December 2026. A management team absorbed by recapitalisation does not open an acquisition programme.
The result: the cost of acquiring an SME client rises at the precise moment the capacity to absorb it falls.
What lowers it, and what does not
What does not lower it: team training, scoring tools, redesign of origination procedures. All useful, and all acting on the processing of a file once it exists. None of these levers reduces the hours spent on files that will never complete.
What does lower it rests on one principle: move the screen upstream of opening the file.
Three concrete forms:
- Qualify before appraising. A short grid applied before opening — existence of usable accounts, clear ownership of the assets offered as security, consistency of the requirement with the real cycle — eliminates in an hour what would cost six weeks to discover. The value of a screen is not in what it lets through, it is in what it avoids opening.
- Receive files already brought into shape. A file arriving with restated accounts, registered and enforceable security and a cash plan tied to the real cycle is not less risky — it is cheaper. And the cost of bringing it into shape is borne by whoever has most to gain: the company.
- Measure, at last. Three figures suffice, and no bank in the zone routinely keeps them: analysis hours per file opened, the share of opened files reaching committee, and the share of non-completions due to an incomplete file rather than a risk refusal.
The question for the next management committee
Over the last twelve months, how many SME files did we open, how many reached committee, and among those that did not, how many failed for want of being completed rather than by risk refusal?
If that last proportion is high, then the problem is neither the market, nor appetite, nor team quality. It is that we open files we cannot finish.
And that, unlike risk, is corrected without additional capital.
Published 14 August 2026. The regulatory facts cited are verified as at that date.
