2degrees says it has pulled out of a Stats NZ programme that uses cellphone data to track population density - which can, in turn, be used to infer how people move around, en masse, during the day.
• For the first time, Stats NZ has collected information about sexual identity
• New Stats NZ data shows more Kiwis were struggling to make ends meet, says the CTU
• Political Roundup: Sats NZ and a census failure of 'epic proportions'
• Statistics NZ Liz MacPherson boss resigns over botched Census
The telco has flagged privacy concerns around the pending commercialisation of the project, and potential mission-creep when that occurs.
And while broadly supportive of the programme and its privacy protections, Spark is giving its customers the chance to opt-out (see below).
However, the population density project has got the thumbs up from the usually querulous Council for Civil Liberties. "We have little objection," chair Thomas Beagle said, noting the steps taken to anonymise the data before it reached Stats NZ.
And Statistics Minister James Shaw told the Herald he thought 2degress' true worry might be potential reputational damage from what some could see as a Big Brother project.
Shaw himself was initially wary, but overcame his qualms after learning more about the project, and two safeguards being put in place: Privacy Commissioner John Edwards reviewing its privacy plan and an independent data ethics panel being given an oversight role.
Edwards told the Herald he saw the population density project as non-controversial, given the data was anonymised, and there was no way to "re-identify" people."
"Re-identifying" is marketing industry jargon for taking anonymised data then "de-anonymising" it to identify individuals by using algorithms to reverse privacy measures, or by cross-matching it with data feeds from other sources.
Compared to similar programmes overseas, the Stats NZ effort could be considered "a model" for good privacy guidelines, Edwards said.
"Maybe 2degrees has the pip because someone else could be commercialising and selling their data," he told the Herald.
Stats NZ's new Data Ventures unit ran a programme between February 2018 and February 2019 that saw tracking code distributed to Spark, Vodafone (and briefly) 2degrees to build a picture of how people move around their suburb or city during the day, based on the changing locations of cellphones. In certain areas, snapshots were taken every hour.
Data Ventures head Drew Broadley told the Herald he hopes 2degrees will ultimately come onboard as he readies the tracking project move from its pilot phase to a full-time ongoing venture.
But 2degrees corporate affairs chief Mat Bolland said no dice.
"We briefly engaged with Data Ventures on a pilot to understand how our network information may be of help to government departments with things like transport and health funding," Bolland told the Herald.
"Network information - very different to customer information - is something we're open to providing for the right outcomes, such as helping with better infrastructure and policy decisions.
"But we won't do this at the expense of our customers' privacy, which we will fiercely guard. We're recently decided not to take the pilot further, because we needed more clarity on things like the potential commercialisation and ongoing requests."
Broadley stresses that the movement data is anonymised. Cellphone numbers are not collected, for example.
The software used for tracking population density simply counts the number of phones in a given area at a set time.
The limited data collection is in part due to the project's privacy guidelines, but also a function of the fact that Stats NZ and its government agency clients simply could not handle the deluge of data that would come from more tightly-targetted tracking.
And it doesn't seem like Data Ventures Crown clients are using the latest data-mining software from Peter Thiel's Palantir. Broadley says a key constraint is that most of them are using Microsoft Excel, which has limits on the number of rows per spreadsheet.
Having closed its February 2018 to February 2019 trial, Broadley says Data Ventures is now consulting with the Data Ethics Advisory Group on a framework for resuming the population density tracking - this time on an ongoing basis. The framework will include privacy safeguards around commercial deals.
The Data Ethics Advisory Group was setup by the government in July with a remit to help government agencies use data more effectively while protecting privacy, and ensure they consider privacy, ethics and human rights amid the rise of artificial intelligence and machine-learning systems.
Its members include academics from AUT and Victoria, the Prime Minister's chief science advisor Prof Juliet Gerrard and, from the private sector, Dr Will Koning - chief data officer of Kantar (the parent company of market research companies Colmar Brunton and TNS).
Broadley stresses that it is only aggregated data, in any case.
"We're counting devices, not people,' he says.
"Think of it like a rugby game when the commentators tell you what the crowd attendance is. They say 'there's 40,000 people here tonight'. They don't say 'Drew Broadley is here tonight'. That's what population density does, it gives the sum total of people in a given suburb at a given hour."
And Shaw also virulently objects to the term "people-tracking" being used to describe the project.
"It's not movement data. It's tracking density over time - from which you can deduce a level of movement but not track it," he says.
The pilot produced some surprising data.
Broadley says it found a couple of ski field areas were actually busier during the summer, with campers, hikers an day-trippers.
"Knowing how many people are going to be in an area at hourly intervals means we'll see the peaks and troughs of tourism for hotspots like Queenstown, know where roads are being used far more frequently than others and help to make sure there'll always be a bus available on your route," he says.
"The Census tells you the residing population, but this gives us a lens into where people work and play - how many go into the Wellington CBD each workday, how many head to Wairarapa each weekend," he says.
Some of the broad trends are obvious, like the wash of people from the suburbs to the CBD in early rush hour.
But Broadley says even when a trend is obvious, there's often no defensible, reliable data you can put on the table if say, you need some hard number to back up a Provincial Growth Fund grant application.
He also says the population density project will help with emergency planning.
"We can also make better preparations for natural disasters, knowing how many people might be in a suburb if there's an earthquake at 2pm on Tuesday or 4am on Saturday."
A virtual census
The population density tracking project is an example of "administrative data", which can be used to bolster information collected during a Census.
Shaw says the plan was to supplement the traditional census approach with administrative data from the 2023 Census.
However, problems with the botched 2018 Census meant this approach had to be pressed into service early, to help plug gaps (although he notes the supplemental administrative data did not include information from the population density tracking project).
Shaw says it's feasible that the traditional Census will be replaced by a "real-time census" based entirely around various administrative data feeds.
"But that's probably three or four census cycles away," he says.
Spark: you can opt out
A spokeswoman for Spark says the telco participated in the Data Ventures pilot via its data analysis unit Qrious.
"Qrious has taken 13 months' of this data and anonymised it to an even greater degree, to further remove any chance of it reflecting anyone's activity or identity," she says.
"Qrious has provided the total count of devices connected to Spark's mobile network in an area unit within an hourly time range to Data Ventures. For example, in X location at 10am on X date, there were 476 mobile devices connected to Spark's mobile network. There is no other demographic or location information provided, and no information is shared about individual users," she said.
She added, "Spark always provides customers the option to opt-out of having their anonymised data provided to Qrious. They can do this by visiting spark.co.nz/help/other/terms/policies/privacy-policy."
A spokeswoman for Vodafone said the telco became involved in the Data Ventures project "in order to better understand the potential public good use of high-level anonymised telecommunications data."
She added, "It's important to note the data is anonymised and aggregated before it gets to Data Ventures and there are no identifying features, just a total number of mobile devices down to a suburb level at any given hour.
"We already use this data to help plan network upgrades, and we believe there are a number of potential public good outcomes from these data insights. For example, the allocation of emergency response resourcing."
Broadley - who had an entrepreneurial background as a co-founder of cloud computing startup Common Ledger - took the reins at the newly-minted Data Ventures in February 2018.
He says the Stats NZ unit's mission is to make data more accessible others, to unlock its value, rather than to be a cash machine for the Crown agency.
However, it still has to pay its way. Data Ventures was founded with $2.6m in funding. It's now spent $2m. The funding was one-off. Once it's one, Data Ventures will have to pay its own way via commercial deals.