Tuesday, March 17, 2015

Now This Is A Pretty Scary Prospect! How Likely Is It That This Will Be Done Well?

This news appeared earlier in the week.

Welfare systems face $1bn upgrade

Phillip Hudson

Joe Kelly

A new hope to welfare system

SOCIAL Services Minister Scott Morrison has declared there would be near-immediate savings and more effective service delivery arising from the replacement of an antediluvian 1980s era Centrelink computer that poses a new impediment to welfare reform.
Federal cabinet is currently considering replacing the system in a move estimated to cost about $1bn with The Australian revealing today that changing a letter for social security payments takes 100 public servants six months at a cost of $500,000.
Centrelink’s “Income Security Integrated System” which handles $100bn worth of payments each year runs on a Model 204 mainframe considered to be too cumbersome to cope with the overhaul of the payments system proposed by welfare reformer Patrick McClure.
Mr Morrison and Human Services Minister Marise Payne are pushing for funding in the May budget to replace the outdated system.
“This is a system that still has manual processing attached to it… And it’s been left to basically wither for many years,” Mr Morrison told Sky News. “Now, the system’s stable, I should stress, but it could run far more efficiently and effectively both for the users and for the government and more cost effectively as well.”
“No doubt it would save money in the short term even I would suggest with the sort of changes you could make immediately.”
Mr Morrison said that when the system was introduced in the 1980s there were about 2.5 million people receiving payments, but that number had increased to just under ten million people today.
Lots more here:
To me it seems clear that the system from the 1980’s has passed its ‘use-by’ date. The issue is surely how this can be achieved without major delays and cost-blow outs of quite disastrous proportions.
Given the track record of such large and in this case pretty complex systems - where one would have to doubt there was even any written down, complete and reliable set of specifications given the shifting specifications of the 30 years from political change - the chance of the needed systems being developed and implemented without issues would have to be well short of 100%.
The two other obvious issues are how to keep the old systems running while the new systems are brought up and the potential economic impact should payments of this scale to the public fail for any more than a few days.
To me the big challenge is going to be to work out how the workload of the present systems can be divided up and then incrementally transferred. Clearly a ‘big bang’ changeover would be just absurdly risky!

Note: the relevance to e-Health is that these systems interact with the other Human Services systems and so they matter to those interested in e-Health!
How do others think a project of this scale and complexity can be derisked and what salary would you see as fair for the project leader?


Trevor3130 said...

Would it be fair to say the "scale and complexity" of driving a new system into Centrelink would be similar to that of importing Cerner into NSW Health and Qld Health? If so, how are they going?
As for payment to the project leader, anyone silly enough to offer to take it on for less than $5m p.a. should not be considered.

Bernard Robertson-Dunn said...

There are two separate issues here. Issue 1 is Model 204. This is a high performance database system. In its time it was state of the art but the technology has not kept pace with other Relational Database Systems (e.g IBM's DB2).

Model 204 itself is not the reason why changing the applications is difficult. That is to do with issue 2.

Issue 2 is the way departments are put under pressure to meet political deadlines. The pollies want new systems and changed systems ASAP.

The result is poorly designed systems where the highest priority is to meet the deadline.

That in itself would not be too much of a problem, if departments were a) funded and b) allowed to go back and re mediate the sloppy designs/coding.

They are not allowed to go back because they are always under pressure to put new stuff in, or to fix the inevitable problems that have arisen from the "get it working at all costs" mentality imposed upon the IT department.

The current Centrelink systems reflect a) the complexities of current and past legislation and b) the poor designs forced upon them by self serving politicians.

Some of the difficulties have been noted by David, I'd add a few.

As a rough, gut feeling estimate, I'd start with two years of strategy, architecture, business process re-engineering and planning. Four years of sourcing a data centre, setting up communications, building the infrastructure, design, coding, unit testing, integration and implementation. Then a full year of training, parallel running and stepwise (if the system has been architected properly) switchover. Add one year for contingency. Seven years as a working hypothesis.

During this time, any changes to welfare legislation will result in significant delays to the project.

There's a very good reason why Centrelink's systems have not been re-platformed in the past: Politicians. Can they keep their hands of legislation for upwards of seven years? That's potentially three general elections.

I don't envy Centrelink.

A project this size has never been attempted in Australia. The Human Services Access card was budgeted at $1.2Billion but that included a lot of registration costs. However the application was pretty simple in comparison.

In Centrelink's case they are talking about re-engineering business processes. How they will implement a new system with new business processes while keeping the old system working is a major problem. Do they build a new system that replicate the old system, cut over and then change it? Do they implement a system that does both the old stuff and new stuff?

My guess is that $1billion for this project is a way too low. I'd estimate $2-3billion.

Anonymous said...

“How do others think a project of this scale and complexity can be de-risked…”

Look no further than the excellent case study of the national PCEHR system. Risk of not making the deadline? Simply turn the system on ready or not. Risk of building the wrong thing? Get Nehta to specify and design it.

Tom Bowden said...

"Income Security Integrated System" makes for a great acronym

Anonymous said...

Centrelink Replacement Application Program?