How Banks Manage Credit Risk

A facility is approved once and then lived with for years — and the living-with is a machine of its own. The economics explain the effort: a bank keeping a few cents a year on each lent dollar cannot recover a lost principal from margin, so the value of noticing trouble early — while the borrower still has options and the bank still has choices — is enormous. Everything on this page is that one sentence, institutionalised.

The machinery has four layers: the routine rhythm (reviews and covenants), the early-warning read, the escalation path (the watchlist), and the balance-sheet absorbers (provisions and capital) standing behind it all. The individual craft the machine runs on — pattern watching, coupling watching, silence watching — is the discipline set out in chapter six of How Bankers Think; this page is where that craft lives inside the institution. At the top of the borrower spectrum the same machinery runs on formalised risk arithmetic — corporate lending and credit risk maps what changes at that scale, and what doesn't.

The backbone is scheduled attention. Covenants — tested quarterly on most working facilities — are less tripwires than thermometers: a debt-service or leverage covenant trending toward its limit is information long before it is a breach. Information undertakings keep the data flowing: management accounts monthly or quarterly, compliance certificates, audited statements annually. And the annual review re-underwrites the whole relationship — the mechanism every product file on this site keeps returning to, because it is the bank's standing right to adjust limits, margins and terms as the borrower changes, in either direction.

The under-appreciated fact about the rhythm is that it is symmetrical. Borrowers experience reviews as scrutiny; they are also the venue where good performance converts into better pricing and looser covenants. A borrower who arrives at review with the pack complete and the story straight is negotiating; one who arrives late and partial is being assessed.

Financial statements are history by the time they arrive. The signals that matter run ahead of them: facility utilization creeping without its usual retreat, an overdraft that stops touching zero, debtor days stretching, reports arriving later each cycle — and the quietest of all, the disclosure that reaches the bank through an accountant or a third party instead of the borrower. Banks wire these into triggers precisely because each one, individually, always has an innocent explanation. The read is in the pattern, against the borrower's own baseline.

This is the institutional home of the watching disciplines: pattern watching holds the baseline, coupling watching asks whether signals are moving together, and silence watching treats what stopped arriving as data. Nothing in this paragraph is compliance; all of it is craft — and it is practised, one file at a time, in the drills.

When the signals accumulate, the file changes lanes. The watchlist moves an exposure from routine to active management: closer reporting, a deeper review cycle, and a named strategy — stabilise, restructure, or exit in an orderly way. Deeper distress moves files to specialist workout teams, whose existence is itself a design decision: the banker who built the relationship is rarely the right person to enforce against it, for the same separation-of-interests reason that shaped the approval journey.

Two things about escalation are chronically misread from outside. It is reversible — credits come off watchlists constantly, and early engagement is what makes that likely. And it prices candour: at every stage, the borrower who brings bad news early keeps options — waivers, restructures, time — that silence forecloses. The drill that works this terrain is Issue 09, where a borrower that still tests clean but has started reporting late matters more than the one with the worst number.

Behind the watching stands the balance sheet's own machinery. Provisions set money aside against expected losses — from the first day of every loan, and more as a specific credit deteriorates — so that defaults are absorbed as a running cost rather than as solvency events. Which produces a behaviour borrowers rarely connect: a deteriorating credit costs the bank real provision expense long before any default, so the bank has a financial reason, not just a prudential one, to engage early. Concentration limits — by sector, geography and single name — stop the book from becoming one bet in disguise, which is why appetite can close against a whole sector while individual businesses in it have changed nothing. And behind everything, capital — the shareholder buffer that makes the whole trade safe to run on depositors' money.

Read together, the machinery answers the question this cluster opened with: why the desk behaves the way it does. The caution, the questions, the watching — none of it is temperament. It is a thin-margin institution, lending other people's money, engineered to notice early and absorb what it fails to notice.