Community

I think "friendly" fraud is a big unaddressed problem here.  Whereby two people collude , money is sent, the sended complains and gets "automatically" compensated and the two parties split the money.  The bar for what a genuine fraud is will be incredibly high.

Great appointment.

BNPL in many markets is going to be regulated in an increasingly aggresive manner.   A photograph of a pay slip is a suicidal way to asses a credit risk.  The US is already reporting on surveys seing 30% of people missing payments.  Open Banking can clearly assist here by giving a lender a much clearer picture of the likelyhood they will get their money back and not letting people sleepwalk into debt they shouldnt take.

That Sheriff Of Nottingham sticks his nose in everywhere !

It's a great solution.  I don't know about the fees but it was pretty straight forward for us to build.  Bodes well.

This is going to age me !  When I started working in payments in the early 1990's everything and I mean everything had to be available 5 9's.  Thats 99.999% uptime, now it seems acceptable to hit 1 9, less than 90%. Have we lost the engineering expertise to build fault tolerant solutions or are we just "bolting" a lot of new technology onto fundamemtally incompatible "Legacy" systems ?

ouch!

You raise a great point regarding synthetic or “real world” data.  This is always a challenge and we tackle this problem daily in a number of ways. 

Highlighting the insufficiencies in your current test approach and solving these issues is just as important as the execution of the tests you already have, automated or not. 

I’d still like to think that by looking at the problem through the lens of 2018 technology you can significantly increase your luck though.

Increasingly complex and aging legacy systems + dash to new technology + outdated testing practices = more stories like this every month.

What we are finding is that it is good for analysing test results, we generate a lot of results and logs. What is often hard to spot is a test that ran fine for 6 months then stopped woking for some reason vs one that has failed then one and only time it ran.  Clearly the former of those two needs closer attention.  AI is very well suited to this.