A bank watches one balance sheet from inside to protect itself. A supervisor watches a whole population from outside to decide where to intervene. The information, the timing, the consequences and the burden of proof all differ — and confusing them produces frameworks that satisfy neither.
In short
- The two frameworks answer different questions for different people. A bank’s indicators exist to trigger its own management and recovery actions; a supervisor’s exist to allocate scarce supervisory attention and to justify intervening in a specific institution.
- The bank has better information and one subject. The supervisor has worse information and the whole population — which buys comparison, the one thing no bank can do for itself.
- The asymmetry that matters most is the burden of proof. A bank acting on its own indicator is managing itself; a supervisor acting on a signal is exercising public authority over a private institution.
- This is why a supervisor cannot simply adopt the bank’s indicator set, and why an authority that does usually ends up with thresholds calibrated to protect an institution rather than to detect one.
- The two systems should still meet, at one place: the supervisor should know where each bank’s own triggers sit, because a bank approaching its own thresholds is itself a supervisory signal.
On this page
Both a bank and its supervisor maintain something called an early warning framework, and the shared vocabulary hides four differences that change everything about how each should be built.
Different questions, different owners#
A bank’s early warning indicators watch one balance sheet from inside, with full access to positions, pipelines and management intent. They exist to trigger the institution’s own actions before its thresholds are breached — a funding plan adjusted, a limit tightened, a recovery option prepared.
A supervisor’s system watches an entire population from outside, with less granular and later information. It exists to decide where to send scarce supervisory attention, and to support acting on one institution rather than another.
The four differences#
- **Information.** The bank knows more about itself than the supervisor ever will, including things that never appear in a return. The supervisor knows something the bank cannot: how this institution compares with its peers on the same definitions at the same date.
- **Timing.** The bank observes continuously. The supervisor observes on a reporting cycle, with a lag that must be stated rather than assumed away.
- **Consequence.** A breached bank indicator starts an internal process. A supervisory signal starts a process that is done *to* an institution, and may become public.
- **Burden of proof.** A bank acting on a weak signal is being prudent. A supervisor acting on a weak signal may have to defend that action to the institution, to its own board, and sometimes before a court.
The fourth is the one that most often surprises. It means a supervisor cannot use the same threshold as a bank even when watching the same ratio, because the two are buying different things with the same number.
The bank and the supervisor may watch the same ratio and still be right to set different thresholds on it, because they are buying different things with the same number.
Where it goes wrong#
Three confusions between the two recur, and each has a recognisable symptom.
- **The supervisor adopts the banks’ indicator set.** It is available, it is already defined, and it produces thresholds calibrated to protect an institution rather than to detect one. The resulting system rarely fires, and when it does the bank has usually already acted.
- **The supervisor ignores the banks’ frameworks entirely.** A bank approaching its own recovery triggers is itself a supervisory signal, and an authority that does not know where those triggers sit has discarded free information.
- **Comparison is left unused.** The supervisor’s structural advantage is the population — the same definitions across every institution at the same date. A system that assesses each bank only against its own history throws that away.
Where the two should meet#
They are separate instruments, but they should not be blind to each other. The productive connection runs in one direction: the supervisor should know where each institution’s own thresholds sit and how close it is to them.
That information is already filed in recovery planning material. It is one of the few inputs a supervisor holds that reflects the institution’s own judgement about its own fragility, and a bank moving toward triggers it set for itself in calm conditions is a stronger signal than the same movement measured against a peer average.
What to do next#
Take one indicator both sides watch — a liquidity coverage measure, a capital ratio, a concentration limit — and write down the threshold the bank uses, the threshold the authority uses, and the reasoning behind each.
If the two numbers are the same, one of them has not been thought through. If the authority cannot state the bank’s number at all, that is the cheaper gap to close first.
Frequently asked
What is the difference between supervisory and bank early warning indicators?
They answer different questions for different people. A bank’s early warning indicators watch one balance sheet from inside, with full access to positions and management intent, and exist to trigger the institution’s own management and recovery actions before its thresholds are breached. A supervisory system watches an entire population from outside, with less granular and later information, and exists to allocate scarce supervisory attention and to support intervening in a specific institution. Four things differ: the information available, the timing of observation, the consequence of a breach, and the burden of proof. The last is decisive — a bank acting on a weak signal is being prudent, while a supervisor acting on one may have to defend that action publicly.
Can a supervisor simply reuse the indicators banks already report?
The definitions can be reused; the thresholds usually cannot. A bank sets its thresholds to protect itself, which means they are calibrated to trigger internal action with room to spare. A supervisor setting the same threshold inherits a level designed for a different purpose, and the resulting system rarely fires — and when it does, the institution has generally acted already, so the signal arrives after the event it was meant to precede. The productive use of the bank’s own framework is different: knowing where each institution’s triggers sit, and how close it is to them, because a bank approaching thresholds it set for itself in calm conditions is a genuine supervisory signal.
What advantage does a supervisor have that a bank does not?
Comparison. A supervisor sees the whole population on the same definitions at the same date, which no individual bank can do for itself. That makes it possible to distinguish an institution moving because the whole system is moving from an institution moving on its own — a distinction that carries most of the early warning value and is unavailable from inside a single balance sheet. It is also the advantage most often left unused: a supervisory system that assesses each bank only against its own history has discarded the one thing the authority is uniquely positioned to observe, and has effectively rebuilt a bank-side framework with worse information.
Should supervisors see the banks’ own recovery triggers?
Yes, and the information is usually already filed in recovery planning material rather than needing to be requested afresh. It is one of the few inputs an authority holds that reflects the institution’s own judgement about its own fragility, formed in calm conditions and approved by its board. A bank moving toward thresholds it set for itself is therefore a stronger signal than the same movement measured against a peer average, because the institution has already conceded that this level matters. An authority that does not know where those triggers sit has discarded free information — and it is among the cheapest gaps in a supervisory early warning system to close.
The programme behind this article
Work through this material with the practitioners who wrote it.
Early Warning Systems for Banking Supervision: Econometrics, AI and SupTech Masterclass
The supervisor’s view of a single bank’s deterioration — econometric, machine-learning and AI models, the supervisory basis for acting on a signal, and the SupTech that puts it in front of the right desk in time.
View the programme →Early Warning Signals in Banks
A structured approach to early warning signals — leading and lagging indicators, baselines and predictive structure, tied to the business-model and cost pressures that erode banks slowly.
View the programme →Early Warning Indicators, Recovery Planning & Resolution Preparedness
The indicators that fire before it is too late — layered early warning indicators, recovery triggers, credible recovery options and genuine resolution preparedness.
View the programme →