“Broken” SME payments industry “makes no sense” in 2019

SME payments are “broken” and the industry’s reliance on old-world processes “staggering”, according to Paul Christensen, CEO and founder of machine learning firm Previse.

“If I walk into Starbucks and ask for a coffee. I get my coffee and Starbucks gets paid instantly,” says Christensen. “If this were a B2B scenario I would walk in, ask for a coffee and then send an invoice in three months. That sounds ridiculous, and yet that is how the entire world of B2B transactions work.”

A survey carried out by Lloyds Bank and the UK small business commissioner Paul Uppal found that 21% of businesses take more than 30 days to pay their invoices. Another 21% take more than 50 days, while 2% don’t pay out for 80 days. Similar research from Marketinvoice, conducted in 2017, valued unpaid invoices that year at £21.1bn, up 60% from 2016. According to the Federation of Small Businesses, late payments cause the death of 50,000 SMEs every year in the UK.

“SME payments is broken,” says Christensen. “I find it stunning that B2B commerce globally, a $125trn industry, works that way. It is staggering. It makes no sense. Of course, historically there are good reasons why large corporates wouldn’t pay until you’d delivered a box of widgets; they’d be checked, the invoice had to make its way to accounts payable, and the payment needed to be greenlit before being sent. But that’s still happening today.”

Invoice factoring can be used by small firms to avoid payment delays and receive payment from lenders. “The average cost of factoring in the UK is 24% APR,” says Christensen. “That means that if I was using factoring, of a £100 bill I would only be receiving around £96. So, the choice is to try and get £100 from a supplier and wait two months for it or try and factor it now and hope I get close to my billed amount. Even then the process is quite painful and involves signing a whole bunch of forms.”

Previse uses machine learning algorithms to predict the probability that an invoice is going to get paid. “Any that look like they might cause a problem will work their way up the queue as normal, but everything else is paid instantly. Before the payee has even looked at the invoice we know if it’s a safe enough risk and will pay it instantly. The SME supplier pays a small fee for the convenience, which covers us and the interest rate of a bank providing the finance.

“Lots of people have tried to solve the problem [of late payments] but no one has used the data,” says Christensen. “It’s only been in the last two or three years, with the advent of machine learning and all the tools it offers that it’s even been possible to build a working solution.”

Turning the oil tanker

Machine learning and data analytics, according to French firm AODocs, can save 15 minutes per invoice when used for processing. A company which processes 1,500 invoices monthly would save 375 hours during that period, it claims. Yet, a major stumbling block to the proliferation of this technology, according to Christensen, is inertia.

“The biggest challenge is overcoming inertia and building momentum,” he says. “These firms are like old oil tankers, while B2C is more akin to driving a group of jet skis around. It’s easy for a firm like Revolut to get thousands on millennials invested because it’s new and adds value. If Revolut had walked into the FTSE 100 company and said: ‘hey we have this cool new thing’ the answer would have been ‘that’s interesting, come back and let’s have another meeting, and refer it to this committee, and then to legal and compliance.”

It’s a challenge to get B2B firms to move, says Christensen. “They’re harder to get to, harder to instigate change with. The larger the organisation – the likes of BP, Shell, HSBC – have upwards of 100,000 employees. Even if this change comes from the top it can be difficult.”

“We’re dealing with one the top 10 largest UK companies, and their CEO said, ‘this is amazing, I want this to happen’ and 18 months later, nothing has happened. In companies so big and cumbersome as these it’s hard for even the group CEO to force change.”

It will take “years” for machine learning and artificial intelligence to become properly integrated within large companies, according to the Previse CEO. “Machine learning is good when you can get a huge amount of data. There are firms across the world sitting on vast amounts of payments data in their enterprise resource planning (ERP) systems right now but no one is touching it. No one is looking at it.”

Related reading