The Victorian government’s central data unit is improving capability in everything from budgeting to tracking diseases. Solutions don’t need to be complicated, says head of analytics Brad Petry — but make sure you plan for implementation from the start.
One of the biggest challenges in the Victorian Centre for Data Insights‘ first couple of years has been to help the rest of the public sector understand what is possible — and what’s worth spending time on.
Initial requests were for things like “compare two spreadsheets and tell me where the differences are” or to put together a dashboard “which ended up getting very few hits”, recalled Brad Petry, head of data analytics at the VCDI, at a recent event at Melbourne University.
So one of the big drivers of the centre’s work since its launch in April 2017 has been to develop a crop of “lighthouse” projects to demonstrate the more useful types of work they could be doing instead.
Which is not to say simple can’t be useful — there’s plenty of low-hanging fruit in government. One of the first such projects was a visualisation of an excel spreadsheet keeping track of the budget bid approval process and funds available.
“It was a bit of a pipe dream idea from one of the deputy secretaries,” explains Petry.
“They always wanted this type of functionality to help the ministers making decisions about the budget have some real-time information about how and where they were spending money and when they were under and when they were over.”
The centre spent two weeks knocking up a simple solution — which the team didn’t think much of at the time — and showed it to the deputy secretary.
“And they looked at it and said, ‘this is one of the best things I’ve ever seen. I’ve never, over all the time I’ve been working in government, seen anything like this and have wanted this for a long time. Just hang on a minute, I’m going to bring the secretary of the Treasury department.’ … All of a sudden, we had 10 senior executives sitting in a room looking at our spreadsheet.”
That tool was used for the most recent budget and continues to be tweaked. But, having such support from senior executives for basic solutions has helped build the groundwork for more complex projects in future.
“It doesn’t always need to be heavy machine learning and unsupervised models to have an impact when it comes to data analytics.”
A central data agency to build capacity in others
The VCDI sits within the Department of Premier and Cabinet, and is led by Julian Hebden, who holds the statutory position of chief data officer.
It was initially set up with a fairly narrow remit to assist the reform agenda emerging out of the Royal Commission into Family Violence. One of the biggest tasks was to improve data sharing between agencies to ensure they had the full picture when making decisions.
“But it very soon became clear we needed to expand the remit to the whole of the Victorian public sector because it was actually very difficult for us to make a delineation between better data for decision-making in the context of family violence and better data-driven decision making across the board,” Petry says.
So the VCDI acts as a hub of data analytics capability within the Victorian government — but ultimately the goal is to help agencies develop the capability to do data work themselves in the future. Commonly, the projects it chooses will help the government do its job better, or save money.
The centre sees itself as a shared service, and there’s no cost to engaging its services — though a couple of times they have used cost-recovery charges for very short deadline jobs.
They work with agencies that are interested in using VCDI’s skillset to identify where pain points and opportunities lie. Sometimes this means taking a hands-on approach, which can bring out novel insights.
“We’ve spent a lot of time and still do at the front line of the services that government offers, going along with police and doing soil tests with people from Land Water and Planning, working with colleagues in the health space and actually going to emergency departments.
“It’s really important for us to understand what those small incremental improvements could be, because that’s when we start to build up something that’s a lot more valuable.”
A diverse staff helps to examine problems from different angles. Petry says there is roughly a 50-50 gender split, as well as a mix of skills and backgrounds. “It’s not all about the data scientists,” he adds.
“Knowing that the data scientist can do the maths and stats is great, but one of the greatest assets in our team is a former data journalist, who knows how to ask the right questions, knows how to find things and investigate them in a way that no-one else in the team does, and can actually communicate in the business language.
“… You don’t have to be a data scientist, I’m not a data scientist, but the way you bring people together and get them to work in a way that’s conducive to change and to effecting an outcome is really what matters. “
Making an impact
The VCDI’s “watershed” moment was Victoria’s response to the Grenfell fire, he says.
The centre was given the job of assisting the taskforce set up to review how much of a problem flammable cladding was for Victoria.
They were able to combine four datasets from private and public sector sources to build a priority inspection list of 1400 buildings out of a total 200,000 across the state.
“Before we came along they were essentially going to send out inspectors as far and wide as they could, at a very large cost, but also with very little appreciation for whether or not they were going to get the right building,” he explains.
“We identified that if this were done in a manual process, it would have taken over 10,000 hours of inspection time to identify these buildings.”
The VCDI has also done some interesting work with mapping tools. In partnership with Google it created a Google Maps-based system for working out which rooftops would be suitable for solar panels, taking into account the direction of the roof and shade cast by trees or buildings, using a combination of aerial imagery and LIDAR data.
The partnership with Google has proceeded under a gratis agreement that has benefited both parties — with the company expanding its corporate social responsibility capabilities to keep staff motivated, and government keen to learn from the tech sector, which holds “all the skills we want”.
The VCDI also contributed to Melbourne’s successful bid to be the third city to trial Uber Air’s flying taxis by building a congestion map of the city for different times of day.
Now the team is working with the Department of Health and Human Services to understand communicable diseases better — where they come from and how many people are affected. This could be applied to any number of public health challenges — thunderstorm asthma, air quality information, and salmonella outbreaks to name a few.
“We’ve been testing out the viability of doing a piece of work where we can identify potential health risks using natural language processing,” he explains.
“So you go into an emergency department, an ED technician takes your details, blood pressure, heart rate, all those types of things, and based on the combination of those, we can potentially identify a type of communicable disease. We’ve been working closely with a team of epidemiologists and doctors at the Department of Health and are looking to roll this out as a trial in regional Victoria in the next couple of months.”
With all this sensitive information moving around, privacy is a big issue. The VCDI works closely with the Office of the Victorian Information Commissioner — located on the same floor — to check they are using data appropriately, and there are “very strict rules” in place to prevent data breaches, including “up to imprisonment for the chief data officer.”
Learning from missteps
But not everything has gone smoothly. A big lesson has been just how important implementation is.
One department they worked with commissioned an analytics project but never acted on it.
“We did a really good piece on analytics work,” Petry says. “But what we didn’t appreciate is that an insight that just sits on the shelf is an insight and nothing else.”
It’s not all bad news — the VCDI has continued to work with the department to identify smaller opportunities, and holds out hope that if they can build an implementation pathway the project will eventually come to fruition.
“We’ve been able to give them some quicker wins. Eight hours a day of saving here, five hours a day of saving there.”
This has led to the development of four criteria the centre uses to assess whether it will be worth working with an agency. The project needs to:
- Have a significant impact;
- Be something the organisation can’t already do itself;
- Have a measurable outcome; and
- Be implementable.
Thinking about implementation up front is important because not only does wasted effort affect staff morale, potential partners will “lose interest” if you’re unable to demonstrate impact.
“Implementation and execution paths need to be assessed up front. I cannot stress this more. My team is very despondent if they spend a long time working on something and someone says: that’s great, thank you, I’ll put it on my shelf and won’t think about it again.”
“The cautionary tale is you always need to think about the change you want to effect, and work backwards from the end,” he explains.
“How are you going to make that change? If I tell you something, or my team tells you something, what will you do about it? That’s a big change for us in terms of the projects we choose to do in our team now.
“Because if people don’t know what change they want to effect, they’re probably not ready for analytics. Because we can do the technology, but the change is really the big thing we’re trying to do.”
Petry says building a new capability inside government has been one of the most rewarding experiences of his career, and urges anyone given a similar opportunity to take it.
But take note: it’s not always an easy journey.
“Building a data capability is complicated, it’s context-specific and it takes time,” he says.
“Don’t expect to have advanced machine learning on day five.”
Join Australia’s public sector leaders at Mandarin Premium. Only $5 a week for a limited time*.
A year ago, after listening to the views of hundreds of Mandarin readers, we launched Mandarin Premium.
Our readers told us they had an appetite for another layer of Mandarin journalism that went well beyond the news – to analyse the nuts and bolts of managing and implementing government policy.
Put simply, our readers wanted something that understood their jobs and would help with it.
And that’s exactly what Mandarin Premium is and does.
We need your support.
For a limited time, sign up to Mandarin Premium for a year at only $5 a week. That’s just $260 a year, and a saving of $180.
Join today by using promo code 5AWEEK at checkout.