EXCLUSIVE - How the Transport Accident Commission in Victoria is using data to drive a client-centric approach
Recently OpenGov had a fascinating discussion with Mr. Bernie Kruger, Business Intelligence and Data Science Lead, from the Transport Accident Commission (TAC), Victoria, Australia on getting the foundations in place for realising the potential of data insights and exploring advanced applications at the same time.
The TAC is a Victorian Government-owned organisation whose role is to promote road safety, improve the State's trauma system and support those who have been injured on the roads, through an insurance scheme. The TAC covers transport accidents directly caused by the driving of a car, motorcycle, bus, train or tram. The TAC can provide support services for people injured in a transport accident as a driver, passenger, pedestrian, motorcyclist, or in some cases, a cyclist.
The TAC can also pay income support while people recover, and in the case of some serious injuries a lump sum payment may also be payable.
Mr. Kruger has a twofold role at TAC. He is the lead for the business intelligence, as well as the data science function. Within the business intelligence space, there are sub-teams. One of them is the business analyst team that is the interface between the business and the business intelligence analysts. Then there is the data management team, which is responsible for all the data manipulation and working out the data ethics. There is a reporting team, and finally a modelling and insights team. The modelling team takes care of model governance and model implementation, after the models have been developed.
“When its not a repeatable piece of work but the business is coming up with a specific question that we need to respond to with a report or a finding, the insights team produces information and report it back to the business,” said Mr. Kruger.
Business intelligence focuses on the operational aspects, while the data science function looks at more of the strategic applications, including advanced areas such as machine learning, text analytics, natural language processing, social network analytics and etc.
Previously, the data science area was separate from the business intelligence. Mr. Kruger was leading data science, and then he was asked to lead business intelligence as well. Though the two have been brought under one umbrella, Mr. Kruger said that they have separate identities. Business intelligence follows the traditional governance processes and methods, while data science is agile and involves quickly developed proofs-of-concept.
Data in TAC’s 2020 strategy
Data insights are a critical enabler of the TAC 2020 strategy and beyond. The agency plans to establish an enterprise-wide approach to translating data. This will allow research and data to be shared across the organisation which will provide insight and inform decision making.
TAC’s 2020 vision has two components. The first is ‘Towards Zero’, which is about reducing the number of lives lost, as well as the number of seriously injured people towards zero. Data is a key enabler here, informing the team’s strategies to address that on four levels: safer cars, safer routes, safer speeds and safer drivers.
In some cases, TAC uses the data directly to analyse and look for specific patterns. And in other cases, the data research for things like accident research is outsourced to the universities. A lot of that data comes from TAC’s partners, such as, Victoria Police and VicRoads.
The second component of the vision is getting people’s lives back on track after an accident. Mr. Kruger said, “How do we respond to their needs, what they need to get back to work, to get healthy and resume their lives. So that is a core area of client-centricity where we use that data to advise us on our clients’ needs.”
Mr. Kruger went on to outline the challenges faced in the use of data.
“The environment in TAC grew organically grew over time and it was mostly reactive as opposed to pro-active. We utilised whatever we had available but as we get more modern and the nature of data changes, the type of data and the speed which we generate data changes, we realised that there are a few inadequacies,” Mr. Kruger said.
One of the primary problems is data governance. Part of data governance is data quality, as well as data ownership. Who owns the data, is it business or is IT? This is further complicated with the move to the cloud. This also leads to weak data lineage and problems with data quality.
Because of problems with data, if the system changes and reports have to be altered, it’s difficult to know the downstream impact on reporting areas. In addition, a lot of information is in people’s heads and it is a challenge to keep documentation up-to-date with all the changes.
Then TAC was using a specific software suite as a ‘Swiss Army Knife’, for everything from loading the data to cleaning, storage etc. Business software was being used for things it was not designed to do. And being married to one software suite like this can be a trap. The total cost of ownership (TCO) can turn out to be very high.
On top of all this, often there is inadequate buy-in from senior management. They do not consider data to be an asset. It is rather seen as an operational tool. The BI (business insights) team/ data scientists are viewed as service providers in many organisations. This kind of culture results in an us vs. them mentality and hampers collaboration.
The value of the data is seldom measured. Organisations are not aware of the monetary value they can attach to certain data of a certain quality. Moreover, there is chronic underinvestment in IT.
Dealing with the challenges
How is TAC addressing these challenges? Firstly, many data management activities are being automated, enabling greater focus on high value analytics, leading to better reporting and better insights.
Everyone is taken along on the data journey, not just the executives or IT. The benefits of data are shared. People are shown what is involved in a day for analyst and the challenges involved in extracting insights from data.
A Data, Reporting & Analytics competency centre has been set up to bring together analytics pockets dispersed throughout the organisation.
Mr. Kruger said that solutions have to be designed together with business. Data scientists should not just get the requirements and design the solution. The business has to be a part of the process. If everyone is not on-board, projects are going to fail.
A working group was created with representatives from various business divisions as well as IT and business intelligence. A review was conducted of the current environment.
Mr. Kruger added, “We also met with consultation companies. We came up with a future state and we did a proper analysis of what is required to support our strategy to 2020 and beyond, so that is sustainable and can adapt to the pace of data. The challenge for us was then of how to bridge that gap. We knew what we needed to do; why we need to do it, but how to do it was a bit of a problem.”
So, options were worked out over a period of 3 months to understand how to get from the current stage to the next stage. This included software options, integration suites vs. individual applications, business intelligence requirements vs. data science, cloud vs. on-premise, enterprise data warehouses vs. data lakes. Skills and capabilities, as well as governance issues were taken into consideration.
Advanced applications (AI, machine learning and predictive analytics)
Often organisations jump on the bandwagon of machine learning, predictive analytics, without building the foundations. It is driven by a fear of falling behind. “We have to do it because everyone else is doing it.”
Other organisations work to get the foundations in place and only then, do they build AI and machine learning on top.
A hybrid approach is followed at TAC. The team understands the importance of getting the foundations in place and is working towards it. At the same time, while data governance and data quality issues are being resolved, in parallel the team explores and experiments with the advanced applications.
Mr. Kruger discussed a few examples. TAC is using machine learning or predictive applications to create client profiles for qualifying them for pre-approved medical services. This eliminates the need for any additional approvals and makes it easier to meet clients’ needs.
Another application being explored is when clients are referred to vocational training service providers; the clients are scored and classified according to levels of need. Then TAC can make baseline predictions regarding the amount of time required for the client to get back to the workforce. That information is passed on to the service provider as a starting point.
TAC also funded research to find early indicators and drivers for mental health problems or persistent pain, as they can significantly hamper the recovery process. In conjunction with that, TAC developed data intelligence tools to pick out behavioural patterns from the massive amount of data they hold on accident, injuries and launch of claims, services TAC is paying for and more, to identify incidences of those indicators, which could be predictive of debilitating pain or mental health problems down the road.
TAC is also in the process of applying text analytics, analysing the clients’ phone calls with TAC and file notes to understand correlations with early indications of these problems. The idea is to predict what the clients are going to require at the earliest stage possible and then respond quickly and efficiently to those needs.
Another trial project Mr. Kruger talked about is the TAC Sofihub Smart Home where data insights are being used to enable more effective long-term care of the patients.
“A lot of our long-term support clients either live in accommodations supplied by us and many of them require 24/7 support and people to look after them. One of the things that we’re trying at the moment is the Sofihub Smart Home, where there are sensors in the house, for example, sound sensors, vibration sensors to see if a client has fallen and can’t get up or have taken their medication etc. The idea is that it will utilise AI to understand what the person’s normal patterns are, when they need to be alerted to take their medication or alert the client or caretaker if there are long periods of inactivity,” he explained.
With AI support, clients can manage their daily lives without having a person present 24/7. This project is still in a trial phase and privacy concerns are being examined.
Projects such as the ones discussed above demonstrate the benefits that can be derived from data to business and to the senior management. They help get buy-in.
Mr. Kruger said, “So, we get the best of both worlds. The fact that we have problems with data quality and we are having to get all the other things in place, does not stop us from doing these applications because we have to explore the mechanisms of doing those. And hopefully it will be incorporated into the newer architecture and platforms”.