Has the software pendulum swung too far?

Has the software pendulum swung too far?

A senior architect told me that corporate governance now means he can only buy products that are already obsolete. I can see his point – if you’re restricted to products that are live at 50+ sites from a company with more than £100 million of revenue you’re not going to get the latest and greatest. The pendulum has swung too far.

Why has this happened? Such purchasing rules make perfect sense for commodity products such as laptops and word processing software. Does it make sense for a payment switch?

There are 3 basic strategies for providing any service:

  • Buy
  • Build
  • Rent

In the payment switch world, all 3 strategies have been successful over the last 25 years.

Many banks bought products such as BASE24, ON/2 and Postillion. This made a lot of sense 25 years ago. Producing high availability hardware and software was hard (although it had been done back to the Apollo era). Picking a product lowered (and shared with other institutions) risk but at some cost. Some pricing models included ongoing licensing over and above the initial licence. This pushed up costs (but ensured that the vendor would be around for the long run).

Other institutions “rolled their own” systems. Around Y2K the single biggest vendor of payment switches was “home grown”! Arguably these are some of the best systems as they focused on a particular customer’s real needs. They are vulnerable to the political accusation that they are “legacy systems” (or “working” as we sometimes say in the trade) and there is a real risk of losing the experts in these systems (perhaps to retirement). They’ve also been hit when a major change comes to the industry such as Y2K and EMV.

Outsourcing has been a tremendously successful model in the Cards space. First Data, TSYS and others have the vast majority of credit card issuing in some markets and are making inroads into debit card switches.

If you’re considering changing your switch platform which strategy should you pick? I’d argue that there are only 2 real options. If the Cards switch is not core to your business ambitions, outsource. Payments is a commodity – it’s no longer bleeding edge, it’s part of the expected infrastructure like electricity, water and gas. You won’t get the freedom to innovate but you will stop having to worry about the switch.

If innovation in payments is key to your business, build and buy. Products will always lag behind the bleeding edge. If your organisation is trying to stay ahead of its competitors you have no choice but to build what you need. The good news is that high availability is much easier to do than it used to be. Software engineering has evolved too. SOA has turned out to be more than just a buzz word – it really allows solutions to be engineered from sub-components. This is where products come in – buying a product to get the “commodity” functionality you need may be cheaper than building your own.

The other big change is in cost of software and hardware.

When I started, writing software was hard. You had to book access to a terminal, type your code in and wait for the compile to finish. When each compile takes 10 minutes typos are expensive. Testing was largely manual. Things have got better and better to the stage where every developer should have their own copy of the “golden” environment, automated regression tests (at unit and system level), source control such as Subversion or Git. Assuming programmers haven’t got dumber they should be more productive – and this is what I’ve found with good frameworks. Running your own team to build what’s special to your organisation shouldn’t cost the earth.

Hardware has got so cheap that (providing you don’t need something unique) it hardly features in the TCO. Commodity x86 servers running Windows or a flavour of Unix get cheaper and more powerful every year. The cost of hardware to run a 1000 tps switch is already well below £100,000 and will continue to get cheaper.

My advice? If your happy with what you’ve got, stick with it. The cost of doing anything including outsourcing will be bigger than any savings over 5 years. If you want to change but don’t see the switch as something special, outsource but if you want to be a “player” roll your own using the best of what’s available in the market. Stick to standard x86 hardware and build something that you’ll be proud to call “legacy” in 25 years time.

1 Comment

Leave a Comment