A public buyer almost always requires the guarantee it receives to be issued by a bank in its own country, often because its regulations demand it and always because it wants to be able to call it without crossing a border. The exporter, for its part, has no relationship with that local bank. A chain forms between the two: the exporter's bank counter-guarantees the local bank, which issues the guarantee to the beneficiary. Three undertakings then coexist, each autonomous from the others, and the risk to read is not the beneficiary's but the chain's own, which always breaks in the same place.
One must first understand that these undertakings are not nested but parallel. The local guarantee is governed by the beneficiary's country's law and its courts; the counter-guarantee is governed by another law, agreed between the two banks. Calling the first triggers calling the second almost automatically, because the counter-guarantee is also payable on first demand: the local bank pays, then reimburses itself on the counter-guarantee, and the exporter's bank debits its customer. That movement happens in a few days, it is mechanical, and none of the links is in a position to examine the merits of the commercial dispute.
That explains the asymmetry of position exporters discover. Contesting the call means acting against the local bank, in its country, which points back to everything known about recourse in a jurisdiction without useful enforcement. Acting at home against one's own bank is easier and almost always futile: it paid in accordance with an autonomous undertaking, it had no business examining the commercial contract, and an injunction not to pay is obtained only in cases of manifest fraud, strictly assessed. The chain is therefore designed so that the money leaves, and so that the discussion happens afterwards, between supplier and buyer.
The place where the chain breaks is almost always the same: mismatched durations. The local guarantee carries an expiry date, often extendable at the beneficiary's request; the counter-guarantee carries another, fixed between the two banks. If the first is extended automatically and the second is not, the local bank finds itself exposed without cover and will demand immediate renewal. The practice known as extend or pay, by which the beneficiary requests an extension while threatening to call if refused, exploits precisely that mismatch: an exporter is then forced to extend an undertaking it believed it had escaped, sometimes several years running, on a completed contract.
A second break, less frequent and heavier, lies in the failure of a link. If the local bank is placed under administration after paying, the counter-guarantee has already operated and the exporter has borne the loss; if it fails before, the beneficiary can maintain the guarantee was not honoured and fall back on the commercial contract, which returns the dispute to its starting point but with sums already blocked. Unwinding the chain to the final risk carrier is therefore not a classroom exercise: it says who bears what in each scenario, and the answer is almost always the exporter.
Insurance intervenes on a precise segment of this chain, and one must know which. What is covered is the abusive or unjustified call of a guarantee, that is, the loss suffered by the exporter when the beneficiary calls without right, and some policies also cover a call resulting from a public decision. The cover does not oppose payment, which will happen in any event: it reimburses afterwards, on proof, and that burden of proof is real. What it never does is stop the mechanism from running, and the exporter that hopes so has the wrong product.
The practical reading amounts to three checks before issue and one discipline during. Before: what is the local guarantee's duration, what is the counter-guarantee's, and is the gap controlled; what event ends the guarantee, a date, an acceptance, a certificate, and who issues that document; and does the commercial contract oblige the beneficiary to return the guarantee on acceptance, failing which it outlives the works. During: demand release as soon as the discharging event occurs, because a guarantee forgotten in a drawer costs money, gets extended, and ends up being called.
A French contractor has delivered a pumping station to a public buyer in Central Africa. A performance guarantee of 2.4 million euros was issued by a local bank, counter-guaranteed by the contractor's French bank. Provisional acceptance was pronounced eighteen months ago; final acceptance, which releases the guarantee under the commercial contract, has never been pronounced, the buyer not convening the commission. The local guarantee expires on April 30, the counter-guarantee on May 15. On April 2, the beneficiary writes that it will call the guarantee if it is not extended by twelve months. What should be done?
The threat is not a bluff and the mechanism proves it right: if the call is made, the local bank will pay on an autonomous undertaking, reimburse itself on the counter-guarantee, and the French bank will debit the contractor within days, with nobody examining the merits. Contesting would mean acting against the local bank in its country, and acting at home against one's own bank would be futile, an injunction not to pay being obtainable only in cases of manifest fraud. The contractor therefore does not have the choice it thinks it has between extending and refusing. Two findings organize what follows. The first is that the guarantee outlives the works through the fault of a clause: final acceptance is what releases it, and the buyer withholds it by not convening the commission, which gives it permanent leverage. Nothing in the contract obliges it to do so by a date, and that is what should have been written before signature. The second is the mismatch of durations, only fifteen days between the two expiries, which leaves insufficient margin and explains why the local bank will demand immediate renewal. What remains to be done is therefore bounded: extend, but for a short period and not twelve months, tying it explicitly to convening the acceptance commission; formally require the buyer to convene it, which dates the refusal and will serve as evidence of abuse; and check whether the policy covers an abusive call, knowing it will not prevent it and will reimburse afterwards, on proof. The lesson holds for the next contract: a guarantee ending on an act only the beneficiary triggers never ends.
- 01The three undertakings in a guarantee chain are parallel and autonomous, not nested: no link examines the merits of the dispute.
- 02A call propagates within days from the local bank to the counter-guarantee and then to the exporter's account, and nothing is designed to stop it.
- 03The chain breaks on mismatched durations, and the extend-or-pay practice exploits precisely that gap.
- 04A guarantee ending on an act only the beneficiary triggers, such as final acceptance, never ends.
- 05Abusive-call cover reimburses after payment and on proof: it never stops the mechanism from running.