Automating high volumes of transactions across different geographies and jurisdictions is yet to be a seamless process. In fact, failed payments remain a common challenge for the industry. A recent research study by Accuity, ‘The
True Cost of Failed Payments’, estimated that in 2020, failed payments costed the global economy $118.5 billion. Against a backdrop of rising consumer demand for fast, simple, and error-free payments, these frictions must be ironed out.
Fortunately, there are a number of options for payment specialists. By optimising straight through processing, for instance, payments can arrive quickly – and at the correct beneficiary account – with minimal interaction from the customer, and no manual
intervention from the organisation processing the payment.
To workshop these issues, payment experts gathered for a Finextra webinar,
‘Improving Straight Through Processing: The true solution for payment failures’, in association with Accuity. The panel looked to identify why so many payments fail in the first place, how payment failure rates vary by type and sector, which technological
innovations can support straight-through processing, as well as the business impact of ignoring failed payments.
The business impact of failed payments
Defined during the webinar by Accuity’s payments specialist Mark Bradbury, as transactions that are rejected by the beneficiary bank, failed payments impact organisations of all stripes – be it in terms of fees, labour, or even lost business.
“The overall cost of managing a failed payment – including investigations, repair costs, reprocessing fees, and exchange rate movements – could easily be over $50 per transaction,” said Momlee Bhattacharjee, SVP, global product lead, domestic payments and
Citi payment outlier detection, Citi.
While this may not sound particularly concerning in isolation, it is important to also consider the average frequency of failed payments. According to Accuity’s global research study, most organisations are subject to a payment failure rate of 5% or less.
However, almost a fifth (18%) report a failure rate of 5-10%. This can represent a considerably high volume of transactions, especially for large institutions. “Every time there’s a failed payment,” said Ronald Wong, head of wholesale payments digital design,
JPMorgan Chase, “it equals dollars lost, clients lost – and a lot of time from the customer service side to try and get clients back.”
Interestingly, operational costs represented the biggest impact of failed payments for 46% of Finextra’s webinar audience. Customer experience, meanwhile, was voted for by 43% of poll participants.
For Wong, the results were no surprise. Failed payments create a negative experience for customers, and therefore increase churn, he explained. Given the level of competition within the payment services sector today, players can no longer afford to deliver
a suboptimal payments service. Unsatisfied customers simply abort the relationship and identify an alternative provider. Indeed, according to Accuity’s 2021 study into payment failures, of the organisations with over 20,000 failed payments per day, 80% reported
customer losses.
When it comes to corporate clients – which still represent a large chunk of a bank’s revenue – “failed payments can also be a liquidity issue,” pointed out Wong. “If you have bill payments coming in, there will be people on the other side expecting the pay
out.” If one or more of these transactions bounce, major reconciliation challenges occur, and cashflow is interrupted. As such, “accounting teams can spend four or five hours a week on reconciliation processes alone,” explained Wong.
If these figures are extrapolated to 12 months, it becomes easy to see why failed payments can also be harmful to staff workload – which, incidentally, was voted as the most serious impact of failed payments by 11% of Finextra’s webinar audience.
Breaking down payment failures
While failed payments are a widespread issue, rates are not consistent across payment types. When it comes to card, for instance, domestic payments are often processed seamlessly – with less than 10% failing, noted Wong. This is because there are very few
variables associated with advanced domestic, automated clearing house (ACH) transactions.
Cross-border payments, meanwhile, could be subject to a slew of different percentages, as a result of the various countries and regulations involved. Real-time payments, on the other hand – “which are part of the Open Banking-type rails – can be validated
extremely effectively,” said Wong.
In some jurisdictions, however, it is extremely challenging to execute a successful payment. For example, around 60% of payments into Brazil bounce, unless an entity is set up, explained Bradbury. This is because, “they ask the customer for a branch code,
which the bank branches do not give out.” Evidently, some institutions need to be more open when it comes to supplying requisite payment information.
Payment failure impact varies across sector and organisation type, too. According to the Accuity report, smaller organisations tend to have higher payment failure rates – particularly in North America and Europe. This can be attributed to fewer resources,
budget constraints, or lack of appropriate technology.
The cost incurred as a result of failed payments also varies by organisation type. The average bank, for instance, spent just over $360,000 on failed payments in 2020 – whereas the average non-bank financial institution spent almost $220,000. This may be
due to the fact that incumbent banks are subject to stringent regulations, and “have compliance breathing down their necks,” suggested Bradbury. Fintechs, meanwhile, are often subject to less regulatory scrutiny, and have a high adoption of application programming
interface (API) technology, which drives down failed payment rates.
All things considered, the entrance of non-bank financial institutions to the market has served to drive down the cost of payments, improve the speed, and upscale the user experience. This competition remains a constant challenge for traditional banks with
legacy infrastructures.
Clearly, there is no single answer to what the number of failed payments should or could be, due to the existence of different payment types, segments, kinds and sizes or organisations, and different Merchant Category Codes (MCCs). Each variable will require
a different solution to raise the percentage of successful transactions.
Common reasons for failed payments
In order to select the correct solution, it is important to first evaluate the reasons why payments fail in the first place. Naturally, these can vary immensely, depending upon the organisation in question, and how their payments data is accessed.
Broadly, however, payments can fail to process as a result of:
- Inaccurate or incomplete information, associated with account numbers or beneficiary details;
- Data entry issues, due to human error or poor reference data and validation tools;
- Data integrity issues;
- Suspicious activity;
- Varying payment methods, and lack of standardisation across the payments ecosystem; and
- Automation deficits and manual intervention.
“One of the main reasons for a failed payment tends to be around the formatting of the file in which the instruction is contained,” said Tristan Blampied, senior product manager, payments, Silicon Valley Bank UK. For instance, a sender may forget to fill
out all the necessary fields, or they may be placed in the wrong order – meaning the IT applications cannot process the instructions, and the transaction bounces.
The confusion around formatting often comes from the fact that different payment types – be it domestic payments or international cross-border payments – require different messaging structures. For example, a payment may need to be sent via “a SWIFT message,
an ISO 20022 Extensible Markup Language (XML) file, or any number of other proprietary or legacy requirements,” noted Blampied.
Missing information is also a common issue – particularly with international payments, which require more information than domestic payments, depending on which countries the payment instructions are passing between. “In some scenarios, it may be mandatory
to include a code that indicates the purpose of the payment,” said Blampied. “Since this isn't necessary information for all types of payments – or even all international payments – it can cause a lot of international payments to be rejected.”
Fat-finger errors, meanwhile, have been a persistent issue for many years. If the International Bank Account Number (IBAN) and Bank Identifier Code (BIC) do not match, for instance, the transaction will not go through. This very issue “comprises almost one
third of all payment failures,” states Accuity’s study.
According to Blampied, there are strong systemic safeguards that can limit these kinds of problems. Reference data, for example, can be used to validate bank information at point of submission – and auto-reject immediately – to prevent delays. Nevertheless,
issues invariably creep through if the reference data is not up to date, resulting in operational overheads, delays, manual processing, and in turn, customer inconvenience.
To improve straight through processing rates, the panel agreed that the industry needs to push for the standardisation of messaging formats across the various payments systems and networks, in order to foster a more cohesive ecosystem. If this is achieved,
automatic and automated data validation will become commonplace, and customers will benefit from an around-the-clock, seamless experience.
The first step in this goal, argued Bhattacharjee, is the sharing of transaction information to address some of the challenges around data integrity and validation, as well as the upfront authentication of data. “We need to come together and achieve straight
through processing, in a real sense of the term,” she said.
Technological adoptions for straight-through processing
According to the Accuity 2021 study, despite respondents’ dissatisfaction toward failed payments, almost 40% of organisations with a 1-5% payment failure rate are not actively implementing changes. The tipping point for action appears to be when failed payments
hit 5% or more.
But what should positive action look like? Broadly speaking, reducing payment failures is about limiting manual intervention – particularly when it comes to, for example, the use of look-up tools – and driving straight through processing capabilities.
According to Blampied, solutions powered by artificial intelligence (AI) and natural language processing (NLP) are particularly effective at improving automation levels, and can undertake complex repairs to messages with formatting inconsistencies or errors:
“AI enables banks and other payment processors to achieve the next level of efficiency. There are limits to how far rule-based processing techniques can go,” he said.
Indeed, if payments processors can implement NLP technology, their IT applications will be able to undertake a much more complex series of repairs. By reading a file or message and assigning a value to each piece of data within it, NLP tools can build a
contextual understanding of the given information and what it is being used for. This learning is then combined with a knowledge base of, in this case, payment information, gathered over time. “The result is an ability to automatically repair payment instructions
with multiple semantic and syntactic errors, without human intervention,” said Blampied. As such, AI and NLP technology promises to considerably reduce rejected payments, and cost risk.
Moving on from AI and NLP, the panel also cited machine learning (ML) as a key technology for driving straight through processing. When linked to a bank or payment processor’s system which the operators use to repair payments, supervised ML tools can analyse
the repairs themselves that operators are making, and create models to ensure future payments received with those same set of errors are rectified by the system in the same manner. “This means straight through processing rates are continually improved over
time,” stated Blampied. “What’s more, as new errors occur, new, and more effective, repair models are created.”
Other honourable technological trends mentioned by the expert speakers included optical character recognition (OCR), for capturing and scanning invoices; robotic process automations, for tasks such as invoice matching; and of course, APIs, which offer a
flexible and iterative response to the challenge of failed payments. However, Bradbury also stressed the importance of retaining some level of human interaction in the payments journey, in order to retain the customers that favour it.
The key takeaway for this stream of conversation was the importance of shifting from being data rich to data intelligent. “While payment rich data, across rails, can streamline reconciliation, an integrated view of data offers financial lenders, real-time
insights into performance,” said Bhattacharjee.
Priorities for payments in 2021
With these learnings in mind, webinar moderator and researcher at Finextra, Jane Cooper, asked the panel what the key objectives for payments specialists should be, in 2021.
Bhattacharjee argued that looking at data, how to use it, and what technologies can be deployed to manipulate it, will be critical to firms’ prosperity in the payments space. By leveraging technologies such AI, NLP, ML – as well as Open Banking capabilities
like APIs – payment processors can achieve customers’ most pressing demands of payments accuracy, speed, low-cost, and security, added Bradbury.
Considering broader, industry-wide goals for the medium-term, Bradbury cited the importance of initiatives such as Confirmation of Payee (CoP) in the UK, which – once fully operational “in three to five years” – will mean transactions are less likely to
bounce as a result of being sent to the wrong account.
However, for pre-validation initiatives like CoP to be a success, the industry will need to reach a consensus on where the line is drawn in terms of what is considered a ‘match’. For instance, should a spelling mistake in the account holder’s surname cause
a payment to be rejected? The debate around this issue continues, but until it is resolved, “we have web-based tools to help us validate payee information,” assured Bradbury.
While today’s validation services can automatically confirm payment data before customers submit their information, the automation of data validation is not yet ubiquitous, and varies considerably by sector and geography.
Achieving 100% error-free payments: The Open Banking dream
Given the economic cost of failed payments, as well as the business impact for individual organisations, it is no longer a question of
if, but how 100% error-free payments services can be achieved. Straight through processing is a critical step in this mission, with AI, NLP and ML technology being increasingly adopted to meet end users’ need for a fast, frictionless experience.
“The dream of 100% will only be achieved once the Open Banking paradigm matures, and we can literally link one bank to another via APIs,” argued Bradbury. “Until then, we will have to work around complex and detailed reference data that comes in different
formats, which we then give to the banks, and the banks build their systems around.”
According to Bhattacharjee, a successful future of strong visibility, security and efficiency in the payments space will require a combination of forward thinking, mindset shifts, innovation, and most importantly, a lot of multilateral collaboration – leveraging
cutting-edge financial technologies. “Anything under 100% should not be acceptable,” she said, “but we still remain a little distance away from that reality.”
To learn more about failed payments,
download this Accuity study, which draws on insights and benchmarking from 200 payment professionals worldwide.