By Julian Bajkowski and Paul Hemsley

A friendly knock on the door every five years from a Census collector used to be the calling card of Australia’s annual statistical stocktake. But the era of thousands hitting suburban footpaths to measure national population trends is rapidly drawing to a close.

The Australian Bureau of Statistics (ABS) has revealed it is now formally out shopping for advice on how to best build its computer systems for the 2016 Census, an event the national numbers agency has previously indicated it anticipates will be a predominantly online affair.

Tender documents released to the market by the ABS this week have called for bids from technology experts to undertake the Census 2016 ICT Capability Review, a process that will independently scrutinize the agency’s own technology strategy and infrastructure to make sure the eCensus doesn’t get stage fright on a night in front of its biggest audience yet.

While the digitisation delivery of government services is becoming increasingly taken for granted, the sheer scale of making a mainly digital Census a reality is, in technology terms, a gargantuan undertaking that would easily challenge the capacity and resilience of even the biggest private sector organisations.

What makes the undertaking such a big ask is that a very large section of the population – literally millions of people – all pile onto a website at approximately the same time to fill in dozens of sensitive and confidential questions.

The sheer volume of traffic hitting ABS’ servers on Census night has previously been compared to the equivalent of a massive denial of service, a load so big that years of planning and testing are required well in advance of the ‘big night in’  so that services stay up and running smoothly.

“The ABS is aiming for a primarily online response for the 2016 Census, making it the Australia’s first digital Census,” the ABS said in its documents.

“While this approach will lead to significant processing efficiencies for the Bureau, it also comes with some significant risks, particularly around data capture and the ICT infrastructure requirements associated with capturing Census information from approximately 6.5 million households in a short enumeration period.”

Just how big those processing efficiencies will be, and the money they will save, will be music to the ears of the conspicuously impecunious Abbott government and Treasurer Joe Hockey.

According to ABS’ own reckoning, the 2011 Census cost “about $440 million, or about $19 per person” with the largest single cost being “$159 million in salaries paid to around 43,000 people to help deliver and collect the Census forms.”

This time around, it’s likely there won’t be many of those temporary employees left.

But to achieve such highly desirable savings (although maybe not so desirable from those benefitting from the part time work) the ABS has to invest substantially in building what might well be the biggest single piece of web computing and communications infrastructure in Australia for what is traditionally a one night stand every five years

That makes buying computing kit outright for the quinquennial sweep a financially questionable exercise   but it’s a big opportunity for cloud and managed online infrastructure services providers seeking a prestigious trophy client. For obvious good governance reasons, experts advising the ABS won’t also be able to bid for the main system build.

There is also very good reason for the ABS to anticipate a massive response to its 2016 offer for online collection.

The 2011 eCensus proved to be a resounding success, surpassing anticipated take-up rates at 33 per cent, a figure that went far beyond the 10 per cent of digitised returns in 2006. The highest uptake was in the ACT on 44 per cent , followed by Western Australia on 35.4 per cent.

The Northern Territory had the lowest take-up rate on 27, per cent, still a high number.

But for Australia’s counters-in-chief, the hordes of Australians going online to complete the eCensus has one distinctly measurable benefit that statisticians as well as technologists can take pride in.

While door knockers often missed people when they were out, the rate of ‘undercount’ or missed survey completions is falling too.

For 2011 the ABS found the net undercount for the 2011 Census was 375,000 people, or 1.7 per cent of the population – meaning that it got to at least 98 per cent of the population.

“In contrast, in 2006 the net undercount was 550,000 people, or 2.7 per cent,” the ABS’ fact sheet on the 2011 Census said.

How far down that figure might come in 2016 must surely be the subject of an internal tipping competition somewhere in ABS headquarters in Belconnen.