Progress in digital trade is usually measured by adoption. How many carriers issue electronic bills of lading, how many jurisdictions have enacted a version of the MLETR, how many banks have joined a platform. Those are reasonable things to count.
Here are two numbers that are rarely put next to each other, and that between them say something the adoption counts do not.
The International Chamber of Commerce puts the share of documentary credit presentations containing at least one discrepancy at around seventy per cent. Published estimates of documents refused on first presentation cluster between sixty and seventy-five per cent, with some regional figures running higher still. And in the ICC Trade Register 2025, built on pooled data from twenty-one global banks and released on 4 November 2025, default rates for the major trade finance products — letters of credit, guarantees, trade loans — remain at historically low levels, below 0.3 per cent overall.
So the instrument fails its own test roughly two times in three, and almost never fails to pay.
Two numbers that should not sit together.
Before drawing anything from that pair, it is worth being exact about what each one measures, because they are not two views of the same quantity and treating them as arithmetic would be wrong.
The discrepancy figure counts presentations. A presentation is discrepant if the examining bank finds anything at all that does not conform to the credit — a date, a spelling, a freight term, a document described in the credit as one thing and presented as another. One flaw is enough. The ICC’s own breakdown of where those flaws sit is instructive: transport documents account for roughly thirty-eight per cent of discrepancies and commercial invoices for about twenty-seven per cent.
The default figure counts credit losses on exposures held by banks. It is the number that supports the long-running argument that trade finance is a low-risk asset class, and on its own terms it is sound. It says nothing about whether an exporter was paid on time, or paid in full, or paid after agreeing to something they did not want to agree to.
The comparison is therefore directional rather than exact. But the direction is not subtle. Whatever the precise figures, the population of presentations that fail the documentary test is enormous, and the population of transactions that end in a genuine failure to pay is tiny. Something is closing that gap, every day, at scale. It is worth knowing what.
The gap is a decision, not a document.
The mechanism is written into the rules, in plain sight, and it is one sentence long.
Under article 16(b) of UCP 600, when an issuing bank determines that a presentation does not comply, it may in its sole judgement approach the applicant for a waiver of the discrepancies. That is the whole of it. The bank is not obliged to ask. The applicant — the buyer — is not obliged to agree. If the buyer agrees, the money moves. If the buyer does not, the documents are refused.
The surrounding articles are all about discipline on the bank’s side, and they are strict. Article 14(b) gives the bank a maximum of five banking days following the day of presentation to examine and decide, and article 16(b) expressly says that going to the applicant for a waiver does not extend that period. Article 16(f) is sharper still: a bank that fails to act in accordance with article 16 is precluded from claiming that the documents do not comply at all. Mishandle the refusal and you have paid.
What the exporter is actually holding
Read those provisions together and the instrument looks different from the way it is usually described.
An exporter is generally told that a letter of credit converts buyer risk into bank risk: ship the goods, present conforming documents, get paid by a bank rather than by a counterparty you cannot chase. That is true, and it is why the instrument survives.
But in the two-in-three case where the presentation is discrepant, what the exporter is holding is a conditional promise whose condition has not been met, plus an option held by the buyer to overlook that. The option has no strike price, no expiry the exporter controls, and no obligation attached. It is exercised or it is not.
Most of the time it is exercised, which is why the default number stays where it does. That is not the same as the exporter being safe. It is the exporter being fine because someone chose.
UCP 600 was the fix. The rate did not move.
This is the part that should stop anyone reaching for a drafting solution.
UCP 600 came into force in 2007, replacing UCP 500, and reducing the level of refusals was among the things the revision was meant to achieve. The examination standard was clarified, the language on data conflicts was rewritten, the notice regime was tightened. It was a serious piece of work by serious practitioners.
Nearly two decades later the discrepancy rate is broadly where it was. The most careful available reading of the evidence is that the introduction of UCP 600 did not improve it and still has not.
A rule change that thorough, followed by no movement at all in the number it targeted, is strong evidence that the number is not measuring something a rulebook can reach. Which raises the obvious question of what it is measuring instead.
Why the rate is stable.
Strict compliance is doing its job
The first answer is uncomfortable and correct: a high discrepancy rate is partly the system working. Documentary credits run on strict compliance precisely because the examining bank has no view of the underlying trade. It cannot know whether the goods were good. All it can do is check the papers against the terms, on their face, and it is supposed to be unforgiving about it, because the moment examination becomes a judgement about commercial reasonableness the instrument stops being an instrument.
Strictness is what makes the promise bankable. It also guarantees that a large share of real-world presentations, produced by real people under time pressure, will fail.
The categories say where the failure lives
The second answer is in the breakdown. Transport documents and commercial invoices are the two largest sources of discrepancy, and they are also the two documents whose content the beneficiary controls least.
The transport document is written by the carrier, from the carrier’s own booking record, in the carrier’s house format. The invoice has to match a description of goods that was drafted by the applicant when the credit was opened, often months earlier, sometimes by someone who has since left. The exporter is being marked on a paper they did not write and a paper they had to copy exactly.
Nobody in the chain is paid to reduce it
The third answer is incentives. The bank earns a discrepancy fee. The buyer gains a lever. The exporter, who bears the cost, is not in the room when the credit is drafted. There is no party to the transaction whose economics improve when the discrepancy rate falls, which is a reliable way to guarantee that it does not.
Short routes make the waiver expensive.
In Asia the timing makes all of this worse, and this is the part that is invisible from a bank’s risk report.
On short intra-Asian routes, the cargo routinely arrives before the paperwork does. The standard workaround is a letter of indemnity: the carrier releases the goods against an undertaking rather than against the original bill of lading, and the paper catches up afterwards.
Now put the waiver conversation inside that window. The buyer is being asked whether to overlook a discrepancy in documents relating to a container that is already in the port, accruing storage, on goods they may already have taken. Every day the conversation continues costs the exporter more and costs the buyer less. A request that begins as an administrative formality ends, often enough, with a price adjustment attached.
The discount that never appears in the loss data
That adjustment is the real cost of the discrepancy rate, and it lands entirely outside the numbers everyone quotes. It is not a default. It is not a claim. It does not appear in a trade register. It appears in the exporter’s margin, and only they ever see it.
What digitisation actually moves.
None of this is an argument against digitising trade documents. It is an argument about which half of the problem digitisation reaches.
It reaches the first half well. When a transport document is generated from the carrier’s own record rather than re-keyed, an entire class of discrepancy simply stops existing: the vessel name cannot be mistyped, because nobody typed it. When examination runs against structured data rather than scanned images, checking is faster, cheaper and more consistent, and the checker’s scarce expertise goes to the cases that actually need judgement. That is a genuine gain and it compounds.
It does not reach the second half at all. The waiver is not a data problem. No amount of verified provenance makes a buyer agree to overlook something they would rather use. Digitisation makes a discrepancy cheaper to find. It does nothing to make one cheaper to forgive.
This is the layer we work on at STANDAGE, and it is worth being straight about the limits of it. The platform frameworks and contractual rulebooks we and others run on reduce the surface area for discrepancy, and they shorten the time between shipment and a complete, checkable file. They do not change who holds the option under article 16(b). Anyone selling an end-to-end digital trade product should be able to say that sentence out loud.
The honest counterweights.
Four things cut against the argument above, and all four are load-bearing.
First, the seventy per cent is an estimate of presentations containing at least one discrepancy, not a measurement of refusals that stuck. A great many discrepancies are trivial, are cured, and are re-presented inside the validity of the credit with no waiver and no drama. The headline number overstates the drama.
Second, the sub-0.3 per cent default rate is a credit loss rate on bank-held exposure. It was never intended as a measure of exporter experience, and reading it as one is a stretch. Delay, fees and negotiated discounts are all invisible to it by design.
Third, the discretion in article 16(b) is a feature before it is a defect. A regime with no waiver mechanism would refuse two-thirds of documentary trade on typographical grounds, which would be an absurd outcome and would kill the instrument inside a year. The waiver is what lets a commercially sound trade complete despite a bad date.
Fourth, UCP 600 binds the bank hard in the other direction. Five banking days, no extension for waiver-seeking, and preclusion under article 16(f) if the refusal is mishandled. An exporter facing a slow or sloppy issuing bank is not without remedy, and practitioners will tell you the preclusion rule has settled more arguments than it has ever had to be litigated over.
The number to ask for.
If you export on documentary terms, you almost certainly know your discrepancy rate, because the bank tells you every time it charges you for one.
Ask for the other two numbers instead. What share of your discrepancies were waived rather than cured. And how many days the average waiver took.
The first number tells you about your paperwork. The second and third tell you about your buyer — how readily they forgive, and how long they are willing to let you wait while they think about it. A buyer who waives in a day is a relationship. A buyer who waives in four, repeatedly, having asked for a small adjustment each time, is running a strategy.
Most exporters have never seen the second and third numbers, and most banks have never been asked for them. Which is worth sitting with, because it means the industry has spent two decades getting better at finding a discrepancy, and has never systematically measured the only thing that determines what a discrepancy costs.
