Let's consider this... A payments issuer handles, say $1.1triilion/year. Maybe $90Bn/month, roughly. Magic, they increase their business 25% in a year. Roughly 1.36trillion/year, maybe $113Bn/month. What would it take to accommodate that increase in volume?
First, customer service. Fraud, complaints, billing questions, you know. Bear in mind, their active cards increased from almost 122 million to over 133 million, about 11 million new cards, and the customers that come with them. There are service costs. It happens.
Second, whatever you were doing the year before, you're processing about 25% more volume. It does not happen by magic. Takes bits. Those need to be kept somewhere afterwards, if only to remember what the heck happened, in case someone asked.
Third, consider that overnight this issuer reconciles accounts. They pay merchants for the charges *they authorized* earlier in the day. And then they wait for the majority of their cardholders to pay them. This capital is also not magic. And it went roughly 25% year over year. Sometimes this is not cheap.
And if you're a merchant that gets paid for these charges in 2-3 days, well, your whine is best served to your acquirer, who holds your money for, well, obvious reasons. Some to wait on the easy fraud that comes from duplicate billings (you pressed back when you should have not), or dissatisfaction over that cold pizza, or your kids are causing trouble on their phones. Or worse,
These were real world numbers up above. Some of you reading this will recognize a 25% increase in data load overnight would make your job more interesting, though you might tell your boss that growth is good, if you can keep up with it. I can haz more budget?
No, costs need not increase 25%, but knowing that typical yearly cards in force increases are 3-5%, you'll be doubling your data every, what, 12-15 years? But that's not what drives IT costs, it's technology, and it costs around $3-6 Billion to build and deploy a new payment reconciliation platform. And that doesn't include costs of reporting, statementing, fraud/risk detection and fraud prevention, customer service, blah blah blah. I'm not intentionally acting as an apologist for payment providers and their fees, but it's a more competitive market than you know, and in a way you are unaware of.
Oh, and you're not managing 10x the payment data for incremental cost increases. You outgrow the SQL tables, then the DB2 tables become unusable, and even. Hadoop, Spark? OLTP is what we're working with here. Cassandra won't perform if you let a merchant dump a million rows of payment data for end of month reporting... And then, when you go past next-day settlement and move to next-hour settlement, your small merchants will be shocked they got paid before lunch. Your reporting will get a little more crazy. You will pay to make that work.