By: Paul Fullam,
Why running two securities messaging standards in parallel has stopped being a bridge and started being a cost
Last November, the payments industry did what the securities industry has avoided for twenty years. They set a date, stuck to it, and switched off the old format. When the CBPR+ coexistence period ended late last year, MT payment instructions were no longer an option for cross-border transactions.
But in securities, there is no such date. ISO 15022 and ISO 20022 have been running side by side for years, covering corporate actions, settlement, reconciliation, and more. The industry continues to use both. Coexistence was meant to be a bridge, but now it feels like everyone has just settled there.
That is the debate I am taking part in on the Standards Forum stage at Sibos this week. My take? Freeze ISO 15022 in time, only touch it if a regulator says so, and shove everything else into 20022.
Coexistence is not free
People like to say that keeping both standards around doesn’t really cost anything. Sure, that would be true if they were totally separate, but they’re not. To keep things interchangeable, 20022 basically gets held back by whatever 15022 can handle. Like, the official corporate action ID? It’s 15 characters in 15022, 35 in 20022, but everyone’s stuck with 15. Same story for decimal places and all that. So if you already paid to move to 20022, congrats – you’re running it with half the cool features turned off.
There are also things that 15022 simply cannot describe. New types of data keep appearing, and if they do not fit into a 15022 tag, they end up as free text. Free text means someone has to read it and re-enter the information. If an important date is hidden several lines down, someone will probably miss it. When operations go wrong, the market does not give you any leeway. I once saw a firm lose millions because two Swift instructions were attached together and the operator missed a page. Every time you put information into free text, you are adding manual work to a process that should be automated.
Why nobody moves
It is common for medium-sized brokers to give the absence of direct revenue as the main reason for not making the change. Since operations are seen as a cost center, the current system is adequate and the fact that the two can be maintained means they can still get the information in a format they are used to. Taking all these points into account, their decision can be understood.
So the incentive has to shift, and honestly, freezing maintenance is the simplest way to make that happen. If all the change requests only go into 20022, the gap between the two just gets bigger every year. Eventually, the head of ops is going to have to walk upstairs and say, ‘Look, we’re sticking with 15022, but now I need to hire people just to read text and pull out data that 20022 would have handed us on a plate.’ That’s the kind of pitch the C-suite actually gets. The only exception is regulation – if a regulator says jump, 15022 still has to be updated so firms can keep up. Otherwise, it just sits there.
Timing matters, because this machinery turns slowly. Change requests being voted on today are aimed at November 2027. Nobody solves a 2029 problem in 2029.
What is arriving next
Tokenisation is the first pressure. Once you can bundle anything into a token, you can pay out anything from it. A payment in kind is an established event type, but the option codes assume the holder is getting cash or stock. If the underlying holds gold, property, or a mixture of the two, then what is being paid, in what unit, and whether it creates a taxable distribution all have to survive the message. At the moment they survive as text. And messaging is only half of it.
Plenty of back office accounting engines cannot store the fractional accuracy that token-based and round-up investing generates, whatever the message indicates.
AI is the second. It is only as good as the data fed into it, and it works best on structured, tagged input. Free text it reads literally, without the market context a 30-year operations veteran brings. It will translate 20022 into 15022 happily enough, and silently drop any element with nowhere to live in the older dictionary. That matters more than it once did, because the people who catch those errors are retiring. AI will handle the simple mandatory events. It will not handle a complex voluntary event in a market with local quirks, because the knowledge to interpret those has never been written down for it to learn from.
Setting the date
None of this means we have to pull the plug on 15022 tomorrow. If firms want to keep getting it, they can – as long as someone’s willing to send it. Some folks will probably still be using it in ten years, just like there are still corners of the industry clinging to ISO 7775. What we really need is to draw a line in the sand and actually stick to it. Payments showed it can be done, even if it hurt. Securities messaging can either look over and see a migration that was painful but finished, or keep running two standards forever and keep asking why the data never gets any better.
I know which conversation I would rather be having at Sibos 2030.
