Why "reduce the backlog" is the wrong goal for infringement reviews

Ask any local laws team across Australia what their infringement review queue looks like and you will likely get a version of the same answer. It is longer than anyone would like. It is worked on by people who would deliver better value to the community doing something else. And it has been like this for a while. "We are working through the backlog," is a common refrain heard from overworked local laws officers. 

Thirty days backlog becomes fifty, becomes seventy 

Nobody decides to have a review backlog. It builds over time. A request comes in and takes a few days to get to. Then a busy fortnight, a big event, an officer on leave, and the few days become a couple of weeks. Turnaround times of 30 days drift to 40. Another busy period at Council and , almost without noticing, your 30 days have become 60 or 70.

Two pressures are leading to longer backlogs, and neither is going away. 

A rising percentage of review requests

The first is increased demands for reviews. Councils are telling us the volume of review requests is rising, and not necessarily because anyone is issuing more infringements. Cost-of-living pressure is changing behaviour: a fine people would once have paid and forgotten is now a fine they contest. For a household watching every dollar, $99 is worth a letter. So more requests arrive, and each one still needs evidence gathered, photographs and notes located, the policy checked, the history considered and a written reason drafted. Done properly, that is anywhere between thirty and forty five minutes of an officer's time. 

The capacity to review can't keep pace

The second is capacity. Traditional answers to a growing queue, for example seconding someone from another department to do the work, is getting harder to do. Rate caps, cost-shifting from state governments and other pressures (for more download our Whitepaper on the Age of Disruption), mean extra headcount is a harder conversation every year. In most councils it is not a conversation the local laws team wins. Reviewing infringements is also a job that few people want to do; local laws officers want to be out in the field interacting with the community and delivering value face to face.

Even when extra capacity is found, it does not fix the problem, it merely delays it.  We recently spoke with a large metropolitan council that had halved a substantial queue over two months by putting one person on it full time. That person did really good work. But it was not a genuine solution. The person was needed elsewhere after a couple of months, and the day that person left the role the queue started climbing again. 

What the wait costs 

In Victoria, a council has 90 days to conduct an internal review and issue a response. If it misses that deadline, the infringement notice is deemed to have been withdrawn and the revenue is simply gone. Money legitimately owed to council gets written off because nobody got to it in time. And as we noted above, the gap between 60 and 90 days is much smaller than it looks.

The less tangible costs hurt too. Residents wait months for an answer to a question they are entitled to have answered, and complaints climb. Officers who should be out doing the job they were hired for are at a desk working through photos, correspondence and records, checking details as small as whether a character on a number plate is a zero or the letter O.

That detail matters more than it seems. In 2020, the Victorian Ombudsman found the City of Melbourne had fined more than 1,200 drivers who had paid for their parking but typed a zero instead of the letter O, two characters that are virtually indistinguishable on Victorian plates. Drivers who appealed were rejected, and the Ombudsman described the council's approach as overly rigid. Read about it here.

That is the hidden cost of a review process under strain. When the queue is long, the pull is to move faster by applying the rules more mechanically, and that is exactly where fair outcomes get lost.

A good review depends on judgement, and judgement depends on officers having the time and the right information in front of them.

Why zero is a better benchmark 

With resource and cost constraints and the real threat of revenue being lost, we need to change the goal. "Working through the backlog" describes real effort accurately, but it is a goal with no finish line. We would put a different benchmark on the table: Zero backlog. This has three things going for it. 

1. It is unambiguous. On any given day you either have an open review queue or you do not. There is no debate about whether the trend is heading the right way. 
2. It is defensible. A council with no outstanding reviews is not losing revenue to a deadline and is not leaving residents waiting. That is a straightforward thing to report to an executive team or a council meeting.
3. It changes the question. "How do we get faster?" leads to working harder at the same process. "Why does this queue exist at all?" leads somewhere more useful - which parts of a review actually need an officer's judgement, and which parts are assembly work that has quietly been dressed up as judgement. 

What Warrnambool did 

Warrnambool City Council took that second question seriously. The deadline is what forced it: with a team of three, they were reaching 70+ days against a 90-day limit, uncomfortably close to writing hundreds of parking fines fines off by default. 

Working with our team at Symphony3, they split the review into the part that doesn't need a person and the part that does:

• Gathering the evidence, reviewing the policy, assembling the file, drafting the basic reasoning: automated.
• Reading the detail, weighing the evidence against the reasoning, and making the call: the officer, every time, with an audit trail behind each decision. 

Officer review time fell from more than 30 minutes to around five. End-to-end turnaround - the wait a resident actually experiences - fell from about 75 days to seven almost immediately. The backlog reached zero within a month of going live and has stayed there. Same team, same headcount. 

It is worth being precise about what changed, because it is the part that gets misread. The officers did not get faster at making decisions. They stopped doing thirty minutes of preparation before they were allowed to make one. 

What it adds up to 

Here is the (very conservative) arithmetic. Our assumptions are transparent so you can substitute your own. 

A manual review takes around 35 minutes of an officer's time. With the assembly work automated it takes around five. That is 30 minutes back on every review. 

A local laws officer costs a council a minimum $60 an hour once superannuation, leave and on-costs are loaded onto salary. So each review is currently consuming about $30 of officer time on work that does not need an officer - finding the photographs, pulling the history, checking the plate, drafting the letter. 

Thirty dollars does not sound like much but it can add up. 

Twenty thousand reviews not a fanciful number for many metropolitan councils. And our

And that is before the revenue. In Victoria a review that passes 90 days is written off. For a council with 800 reviews already sitting in the queue, that is about $79K at immediate risk using the conservative calculations above.

None of this is a case for fewer officers. It is a case for the officers you have spending their hours on the job they were hired for.

But it is also the sum a finance manager will do anyway, so it is better done in the open. 

The part people ask about first 

The first question in almost every conversation about the Infringement Assistant is not about the technology. It's usually about who is actually making the decision.

The officer is. Everytime. And that is a design choice: A human in- the loop everytime. We have we deliberately left excess friction in the process and have chosen (with Warrnambool staff) not to automate further.

The system assembles a finished file - evidence gathered, policy applied, reasoning drafted - and then it stops and waits. Nothing is issued, withdrawn or waived without a person approving it, and a person can halt the process at any stage. 

It is why we call it an assistant rather than an agent. That distinction matters to councils being asked to move quickly on technology while remaining accountable for every decision they make about a resident. Those two obligations can sit together, but only if the human stays in the loop by design. 

The questions that follow

 The questions that follow are of course fair and to be expected. Where does the data sit?  Who can see it?

We run the solution entirely in Australia on Amazon Bedrock. Council data is never used to train the underlying models. 

The part that surprises people

And the part that surprises people is that, while we of course use some AI, most of the heavy lifting is the automation/integration at the front end. THis is where we use our SmartGlue integration platform to collate the evidence from the systems it already lives in and putting it in one place. That is where most of the thirty minutes actually goes when officers do it manually. 

Hear it from Warrnambool 

On Thursday 22 October we are sitting down with Warrnambool City Council to go through the whole story- how the backlog built, why they set the benchmark at zero, how they got there, and what was harder than they expected. 12:00pm  - 1:00pm AEDT, 45 - 60 minutes, and the recording goes to everyone who registers. 

If you want to know what zero looks like for your council before then, we will run a backlog assessment using our calculator tool. Bring your current queue size and average review time and we will show you the numbers.