swift payments

The difference between SWIFT's global payments innovation (gpi) and a fintech start-up doing a proof of concept is that the former is now being piloted by banks using their production systems.

Amid the headline-grabbing news about blockchains fixing, well... everything, SWIFT has been working towards a go-live date early next year for gpi (estimated at less than six months since the announcement about the successful pilot results at Sibos in September this year). It promises to "significantly improve the customer experience in correspondent banking by increasing the speed, transparency and end-to-end tracking of cross-border payments".

As soon as you are making payments across borders where a currency exchange takes place, things slow down to as much as three working days, and the payment itself vanishes into a black hole, as it moves from one account to another.

It should be said that SWIFT, particularly its Innotribe division, has been looking closely at blockchain and the technology is under evaluation in the gpi roadmap, also next year.

However, for now, Wim Raymaekers, SWIFT's global head of the banking market, makes clear the distinction between the gpi pilot and the sort of thing blockchainers announce on a daily basis.

He said: "The gpi is different from some proof of concepts out there. You know, where they say we sent a payment in 3.5 seconds from bank A to bank B; that is typically a demo; there are no real bank back-office systems behind it."

"So here we took the complete opposite approach. Banks are actually coding this gpi into their systems; they are automating the gpi flows. That's what they are testing now in pilot on their production systems."

"Some banks will do what is called 'a penny test'. They will send a few pennies from bank A to bank B and back, but using a real production environment. Some other banks are even ready to start using some real corporate payments and do them via gpi, as part of their pilot testing."


One of the chief components of gpi is to provide end-to-end payments tracking. This will be managed by a "Tracker database in the cloud", securely hosted at SWIFT. Here is a good example of the sort of data management banks feel comfortable with. Privacy of data is a serious stumbling block for all but the most lightweight financial use cases for blockchain. Notwithstanding the game theoretical attributes of these "trust machines", banks are nervous about sharing any data on distributed ledgers.

"The tracker is hosted at SWIFT. It will be accessible using SWIFT technology, using secure APIs. And it means that for the first time in 40 years you will have a single dashboard view to be able to track your payment across banks from start to finish."

Raymaekers said that SWIFT assets come with their own very strong governance framework and identity management system. He added that SWIFT's expertise on standards is also vital. "You can talk about a blockchain PoC for liquidity for nostro/vostro, but which items do you report on? There must be an agreement on how you create the ledger. What transactions do you put on there? We have a lot of experience in standardisation and business rules, and the data model is very important. DLT is just a technology to disseminate things but you must agree what is being disseminated and in what format."

In terms of speeding things up, a community of global banks have agreed new business rules for cross-border payments and will use the existing SWIFT infrastructure to process them on the same day. Raymaekers pointed out that corporates are not always asking for real-time cross-border payments, but what they do require is a high degree of certainty about them. No payment leaves a corporate without a purchase order, just from an audit perspective. The terms of the purchase order state when to make the payment. The terms are usually in 30 days or 60 days, so the payment can be scheduled.

"When corporates make a payment, it must be with certainty. They must know when it gets there. They need to have transparency to know the payment has been made and where it is, and on fees. That certainty is what's important here. There are some urgent payments yes, but the majority of the payments are scheduled."


A slightly alarming development in the world of banking blockchains was the news that some big banks had chosen not to renew their membership in the R3 consortium. IBTimes asked Raymaekers how he reads this.

"Again, SWIFT's approach is to look for live adoption. That's the state you want to get to, that is what we are doing with gpi. Banks are automating and changing their systems to embrace and cater for gpi.

"So proof of concepts are good but the real objective should be more like a proof of pilot readiness. You see the difference. The gpi pilot has been proven. There are banks that have now done this; it works. With proving a concept, you are just at the beginning. Then you could stop and say, 'ah that was interesting' and you return to your day job.

"If purely a PoC, you are putting investment into a concept, but you don't get a return yet so it's just costing money. Sooner or later you are going to ask, how much more money are we going to continue to put into this?"


Previously, cross-border payments blockchain start-up Ripple has said it would be happy to work with SWIFT; Ripple's technology would do the payment settlement and SWIFT would be the messaging layer in their application stack, and also impart considerable clout with its 11,000 members in the banking community.

"They proposed that," noted Raymaekers. "This is because in Ripple you could potentially consider settlement between banks. But for the exchange of personal data, that must be confidential. So you must convey the payment instruction, the personal data over a secure channel. That part can be over SWIFT – like in the MT103 – and then the settlement of the funds between banks to facilitate that money transfer could be done using Ripple."

So why has SWIFT not teamed with Ripple up to test this out?

"In the past, we asked banks if they were willing to consider this and to let us know," replied Raymaekers. "We just haven't had any ask."