Reducing the Harm Caused by Distracted Drivers

Challenges to measuring contributions to road safety

There are many behavioural, environmental and technological factors that influence road safety.

Attributing a reduction in road deaths and serious injuries to a specific program is inherently difficult. An increase in overall road trauma does not mean the program is having no effect. 

But this difficulty was known from the start of the program, and one of the program’s main expected benefits is a reduction in road accidents, serious injuries and deaths. The department has not yet been able to measure this.

Contributing to road safety

The department cannot yet show that the program is contributing to reducing accidents, serious injuries and deaths. 

In 2022 the department prepared a business case that identified benefits it expected the program to deliver. 

Figure 2: Surveyed drivers and passengers reporting using a mobile phone and improper seatbelt use

Note: Figures show the percentage of drivers and passengers of the total surveyed by the Transport Accident Commission.
Source: VAGO, based on Transport Accident Commission data.

Evaluation timing

The 2022 business case said that the department would confirm whether the program had reduced crashes, injuries and deaths within 12 to 24 months of enforcement starting (that is, by late 2024 or 2025). If this had happened, the impact of the program on accidents, injuries and deaths would have been evaluated before the program’s initial funding ended in July 2025.

The department commissioned an external agency to evaluate the program in February 2025. 

The department has shown good practice by appointing an independent evaluator. However, the evaluation was not commissioned until well into the initial funding period, and later than the department planned in its 2022 business case. Evaluation of whether the program has reduced accidents, injuries and deaths has not yet been done. 

Progress to date 

The department commissioned the external evaluator to complete the evaluation in 2 phases. 

Phase 1 of the evaluation is complete. It looked at the number and patterns of infringements since before the program started, and re-offending rates. It found that the program:

• is effective at catching offences on a large scale, given the volume of images and checks it can process
• is less effective at changing individual offenders’ behaviour compared to receiving an infringement directly from a police officer or by going to court 
• may not achieve long term changes in driver and passenger behaviour by itself. The evaluation report suggested the department could increase the program’s effectiveness by combining it with: 
  • other types of enforcement, for example police officers on highway patrol
  • more public education and awareness campaigns. 

Phase 2 of the evaluation is due to finish in May 2027. It will assess the program’s overall impact, including: 

• whether cameras are deployed fairly across regional and metropolitan areas
• how road safety results compare to initial forecasts 
• if the program has reduced the number of road accidents, serious injuries and deaths.

Missing baseline data

The department could have established an evaluation approach from the start of the program using the measures and targets from its original business case. This would have allowed it to gather baseline data to report against later. 

But it did not do this. Instead, it developed an evaluation approach while the program was already running and commissioned the evaluation 2 years into the program.

The department’s evaluator has baseline data about infringements that goes back to 2018. But other critical baseline data is missing. This impacts the quality of the evaluation. 

Complete and accurate baseline data to evaluate if road accidents, serious injuries and deaths are caused by distracted driving or not wearing a seatbelt does not currently exist. The department has acknowledged this.

The 2022 business case

The department’s 2022 business case for the program:

• identifies the benefits it expected the program to deliver, including behavioural changes and road safety benefits
• includes measures and targets to track progress against these benefits
• says the department would work with a research organisation to confirm reductions in road deaths and accident costs within 12 to 24 months after the program started.

This set an expectation that the program’s contribution to road safety would be measurable, including the program’s contribution to reducing serious injuries and deaths.

Although the business case did not quantify the expected reductions, research referenced in the business case suggested that large-scale rollout of cameras could avoid up to 95 accidents per year. 

The research also indicated that the program could result in up to $21 million per year in cost savings relating to accidents. The department has not identified cost savings relating to accidents as a benefit.

The new plan

In 2025 the department decided that the program’s benefits, measures and targets from the 2022 business case were no longer appropriate. 

It told us this was because of changes to the program, including the government’s decision to extend the program in late 2024. 

In January 2026 it released the new plan for the program, which describes: 

• the benefits the department expects the program to achieve that will contribute to road safety
• how and when the department will measure if the benefits are being achieved.

The department’s external evaluator will review and refine the new plan to confirm that its benefits are:

• measurable and reportable (for example, relevant data is available)
• ready for reporting from July 2026.

Until it has done this, the department will not know if baseline data exists to measure progress against the new plan’s benefits. The department has acknowledged this. 

As of April 2026, the department is still refining the new plan. It is due to complete this in June 2026. But this means the department does not have a final set of benefits, measures and targets to assess how much the program is contributing to road safety.

Comparing the new plan to the 2022 business case

The department has changed the benefits it expects the program to deliver and the department’s new plan has less benefits than the 2022 business case. 

These changes to the program’s benefits mean that:

• the department is not assessing as much of the program’s contribution to road safety than it originally planned
• it is difficult for the department to measure the program’s contribution to road safety between 2023 (when it started) and 2026 (when the benefits measures changed) because some of the 2022 business case measures will no longer be used
• the department has a limited view of how well the program is delivering its expected benefits until the evaluator’s results are available and a baseline is established.

While there are fewer benefits in the new plan, the department has identified benefits that align with the program’s broader objectives and focus on its contribution to road safety. For example, the new plan includes benefits such as ‘A reduction in distracted driving and seatbelt non compliance related fatalist and serious injuries’ and ‘Improvement in compliance with driver seatbelt and passenger seatbelt use’. This means the new plan’s benefits are useful to assess the program’s contribution to road safety, if the department can measure and report against them.

Operational data reporting

The department has operational data that measures the program’s activities, including:

• KPI data for Jenoptik’s activities
• the number of infringements issued through the program, including where and when
• fines revenue generated by the program.

The department reports some of this data to external stakeholders, including other agencies involved in delivering the Victorian Road Safety Strategy 2021–2030. Some data is also published quarterly on the Cameras Save Lives website.

The department’s internal reporting has focused on how Jenoptik is performing against its KPIs. This can show how the program is being delivered, for example whether cameras are operating for the required number of hours per month. 

But this operational data is not designed to give a full picture of how the program is performing or contributing to road safety.  

New governance committee and reporting

The department is reporting to its governance committee, but reporting has not been consistent and has limitations.

When the program was first established, it was funded from July 2023 to June 2025. In December 2024 the government extended the program to June 2030. To support the program over this extension, the department created a new governance committee in August 2025. 

Its role is to oversee the program and make sure that its benefits are being delivered. The committee has representatives from the department, Victoria Police and other road safety stakeholders such as the Traffic Accident Commission.

In March 2026, the department reported to the committee against operational metrics, including the number of:

• camera hours delivered
• infringements issued.

It also reported against one of its performance measures around road-user behaviour change and compliance.

The department’s reporting in March 2026 shows improvements from its previous reports to the committee, which did not include this detail. But: 

• the March 2026 report was the first that contained this type of information, so we cannot confirm this is going to be a regular, standard approach
• there are still gaps and limitations in the department’s reporting because of the known issues around baseline data and incomplete targets. 

This lack of consistent reporting makes it difficult for the department and other stakeholders to fully understand how the program is contributing to road safety.

Criteria for camera sites

Victoria Police has 4 criteria to assess, prioritise and identify locations for camera sites, as Figure 3 shows. A site can meet more than one criteria. The department and Victoria Police developed the criteria before the program started.

Figure 3: The 4 criteria to identify camera sites

Source: VAGO, based on Victoria Police's information.

Issues with the criteria

There are issues with the criteria that make it difficult for Victoria Police to assess a location or gather the evidence needed to recommend it as a camera site. 

The issues are:

• Available data cannot always link accidents to specific behaviours (such as distracted driving and not wearing a seatbelt), which is needed for criterion 1.
• Victoria Police does not know the type of road environment or physical characteristics that increase the likelihood of distracted driving or seatbelt offences. This means a location cannot be assessed against criterion 4, which has never been used.
• When the department initially developed the criteria it did not outline what evidence is needed to support Victoria Police’s decision-making. Without this guidance, Victoria Police does not have a standard of evidence to help it decide if the criteria apply to potential sites.

Because of these issues, Victoria Police cannot show it is prioritising camera sites based on the likelihood or risk of distracted driving and seatbelt offenses taking place.

Issues with how Victoria Police uses and records evidence 

When Victoria Police identifies a camera site, it records the site in its rostering system with the criteria it used to make its decision. But Victoria Police does not always include enough information to show why it recommended a site or how the site meets the criteria.

Between July 2023 and September 2025, Victoria Police marked 90 per cent of its recommended camera sites as meeting criterion 1. 

This criterion requires Victoria Police to check if a location has a history of accidents linked to distracted driving or seatbelt offences. Victoria Police uses mapping software to find locations with a history of road accidents to do this. It then records this information in the rostering system. But: 

• out of 33 sites we sampled, 15 did not have accident records against them. This means Victoria Police cannot show it always considered a site’s accident history before recommending it
• available data does not always directly attribute accidents to distracted driving or seatbelt offences. This means Victoria Police may not have enough evidence to show a site has a history of these offences.

We also found 8 cases where Victoria Police did not correctly classify a site under the criteria. For example, it marked a site as meeting criterion 2 (a referral from the public) when it was referred by a police officer (which is criterion 3).

This inaccurate record-keeping means that Victoria Police's decisions may not always be based on sound evidence. 

Victoria Police told us that it reviewed how it documents it camera site decisions in 2025. But we still found that evidence was not sufficient for the sites we looked at.

Results not used to improve site selection

Victoria Police is not using data from the program to plan for or improve how it selects sites.

Victoria Police does not review camera information to get insights about sites or its criteria. For example, it does not know if there are road conditions or characteristics that make a particular site more or less effective at finding distracted driving or seatbelt offences. This particularly affects criterion 4, which asks for this information to assess a site. 

This is a missed opportunity for Victoria Police to collect and use information to improve its criteria and make sure that it is consistently selecting sites.

Camera rostering 

When Victoria Police has identified a potential camera site, it passes it on to Jenoptik to assess if the site is practically suitable. This includes whether: 

• it is safe for staff to put up and take down the camera on the site
• there is any suitable parking space nearby. 

Victoria Police records all sites onto the master list in the rostering software. It uses the master list to create a monthly roster of ‘sessions’ for Jenoptik to set up cameras.

If Jenoptik finds that a session on the monthly roster is not feasible (for example, because of roadworks or bad weather) it notifies Victoria Police that it cannot follow the roster and uses another site on the master list instead.

To avoid delays to camera deployment, Jenoptik can swap unsuitable sites for another site without waiting for approval from Victoria Police. Victoria Police approves these changes later, checking that Jenoptik has relocated to a valid site that is on the master list.

Verifying images

Jenoptik’s process for verifying images is identifying offences effectively. 

Jenoptik makes sure multiple staff independently verify images through different stages of its process. This is to cross-check decisions and confirm that images have enough detail and clarity to be enough evidence of a potential offence. 

Victoria Police then makes the final decision about whether an offence has been committed. If it has, Victoria Police issues an infringement.

Figure 4 shows this process.

Between July 2023 and September 2025, the program’s cameras photographed 31,437,979 vehicles. Of these, Victoria Police issued 188,493 infringements for distracted driving and seatbelt offences.

Figure 5 shows the proportion of vehicles photographed that progressed through each stage of the process.

Low-confidence images

If Jenoptik staff are not sure whether an image shows an offence being committed, they can reject the image during the verification process. These images are known as low-confidence images. 

An image may be low-confidence when, for example, it does not clearly show if a driver is holding a mobile device or something else. Figure 6 shows an image Jenoptik gave us as an example of an object that could be a mistaken for a mobile device. 

This approach helps to make sure Jenoptik does not incorrectly say an offence may have been committed

Figure 6: Example of an object that could be mistaken for a mobile device

Source: Jenoptik.

Issuing infringements

Victoria Police decides whether to issue an infringement. Its approach to decision-making is effective and reduces the risk of it issuing an incorrect infringement. 

Victoria Police uses the same team, procedures and processes to issue infringements for: 

• distracted driving and seatbelt offences 
• other road safety camera infringements. 

This makes sure that Victoria Police is consistent across all its road camera safety programs.

Staff who make decisions on whether to issue infringements get assistance from their supervisors for images they are unsure about. 

People who receive an infringement can contest it in court. In these cases, Victoria Police carries the cost and risk of legal proceedings. This gives Victoria Police an incentive to make sure its decision-making is accurate when assessing whether to issue an infringement. 

The regulations' privacy requirements breached

The department routinely checks images from Jenoptik’s verification process as part of its contract and performance monitoring. 

The regulations allow Jenoptik and the department to keep images for up to 6 months for those purposes (regulation 40C). 

We found cases where the department breached the regulations:

• The department shares images with Jenoptik as part of its performance management process. For example, the department emails rejected images to Jenoptik to check how it made its decisions. We found examples of documents and emails with images that the department had kept for longer than 6 months. The department has since deleted these images. But we have not seen evidence that it has processes to routinely delete images from its email databases or internal filing systems.
• The department has a procedure to check how Jenoptik’s AI is performing. It retrieves images from the cameras and reviews them for accuracy. The procedure says when the department uses the images to check the AI’s performance, it must delete them within 6 months. But: 
  • the procedure also allows the department to keep images indefinitely when it uses them to improve the AI’s performance. This creates an exception that results in the department breaching regulation 40C
  • the department did not fully follow the procedure and held some images it was using to check the AI’s performance for longer than 6 months. The department has since deleted these images.

The department told us that it had kept images relating to the AI’s performance for longer than allowed because it needed them to comply with an information request in a different VAGO audit. That audit was assessing the department's processes for validating fines revenue. 

But the department did not inform VAGO that the information request would result in it holding images for longer than 6 months. The department told us that it was not aware it could decline VAGO’s request. But we have not seen any evidence that the department considered regulation 40C requirements when responding to VAGO’s request.

These examples show that the department needs to strengthen its administrative processes to make sure it meets regulatory requirements on how it retains images.

The department does not validate all of Jenoptik’s KPIs 

The department uses 21 KPIs to make sure Jenoptik is delivering the program as expected. Jenoptik must meet targets for each of them. Figure 7 shows the types of KPIs used.

Figure 7: Program KPIs

Note: There are 22 KPIs in the contract between the department and Jenoptik. But KPI 1.2 measures the performance of fixed-point cameras, which are not currently part of the program. The department does not use this KPI.
Source: VAGO, based on the contract.

The department cannot be sure that results against all its KPIs are accurate. This is because for some KPIs, it relies on Jenoptik to self-report results and does not verify them. 

The department receives monthly reports on all of Jenoptik’s KPIs. It checks Jenoptik is meeting them differently, depending on the KPI.

This shows that the department needs to strengthen its administrative controls over Jenoptik’s performance.

AI performance improvement

If the AI can effectively identify and reject images that do not show an offence being committed, fewer images will need human review. 

As human review is time and labour-intensive, this means upgrading the AI can improve the program’s efficiency. 

The department uses a third-party company to test changes to the AI and see if it can improve performance. This is good practice as it separates AI testing from Jenoptik, which operates the AI. 

In March 2024, the department approved a change to the AI software as a result of the company's testing. Jenoptik then trialled the changes with 2 cameras. But the department does not have records to show what happened after May 2024. 

This lack of record-keeping means the department could not be certain of the version of the AI software that was deployed from March 2024.

This situation was remedied in December 2025, when the department approved a new version of the AI software as a result of this testing. Jenoptik rolled out the software to all cameras and the department monitored these changes. 

It is too soon to assess if these changes improved the performance of the AI as intended.