Dear IRD: how to shave $1b from your $1.5b software spendup

GUEST COMMENT: To anyone in IT, $1.5 billion an obscene amount of money to spend on an software project. It reeks of global consulting firms winning the business and then rapidly hiring a bunch of grads and putting them up in hotels for years.


The New Zealand Government has recently agreed to spend $1.5 billion to redo the New Zealand tax system.

To anyone in IT this is an obscene amount of money to spend on any software project.

From the outside it seems like a slow moving train crash reminiscent of earlier Big Bang projects that always blow out if they are ever delivered.

It reeks of global consulting firms winning the business and then rapidly hiring a bunch of grads and putting them up in hotels for years.

It's just not smart.

A $1.5 billion  injection into local service companies, that are world class, would grow an industry. Government spending of this magnitude should see numerous other benefits.

It's easy to say nothing but the fact is government officials have no idea what's reasonable. The companies with the budgets to win these projects are the people officials meet.

To a well meaning amateur $1.5 billion seems a massive amount of money for a relatively moderate volume transaction system.

We're a market of 4 million  people and 400,000 businesses, so it's just not that big. Many SaaS [software as a service] companies are already a good portion of those transaction levels at a fraction of the cost by using commodity, high performance, technologies.

Xero has spent around only $80 million getting to where it is today. Even if IRD was 10x Xero (it's not) why isn't $800 million a reasonable number?

Doing massive projects delivered over multiple years is counter to the interactive, early and frequent delivery approach modern software projects use to de-risk and deliver value early.

But rather than just criticise here's some practical suggestions I'd offer to to see if we can save $500 million to $1 billion in spend.

1. Start from the customer and work in, replacing the edges. Identify the key external interactions and publish those as web services.  Get the messages into a commodity systems and then connect these systems to the core FIRST servers. That will take load off, allow quick wins and lots of options.  As the core engine is surrounded it can be gradually replaced. A GST Return WebService would be an ideal place to start.

2. Don't build the retail tax front end. Just publish the rules and invest in just the very core system. Let the private sector invest in the layer customers interact with. Certify providers that they met the requirements.  Payroll software pretty much works like that now. That offloads the investment to the private sector who are happy to build.

3. Go out to the NZ service companies and get them to stand up a consortium and carve up the opportunity themselves and put in place the appropriate governance structure. Give them the challenge to save $500 million on a fixed-fee basis and transfer project risk to the private consortium.

4. Appoint an independent board of systems experts to review the project and provide ongoing governance over it.

There's some quick starters. I'm sure others have ideas on how to save hundreds of millions of dollars. Welcome your comments below.

Rod Dury is CEO of Xero.

Login in or Register to view & post comments