I work night fill at a supermarket from 6 pm to midnight, and on some mornings I come back at 7 am for stocktake. During night fill I find what people leave behind. Drink bottles, lolly bags and chocolate wrappers all turn up empty on random shelves, eaten in the store and never paid for. At stocktake I see what is gone. Batteries and expensive items that normally have security tags come up short, with the tags cut off.

Woolworths plans to install new security gates at store entrances, along with a dedicated security guard at the exit to prevent shoplifting. Woolworths says theft is one of the main drivers of violence and aggression towards its staff (Rowley, 2026). Shoppers have complained about the gates. Parents say the low gates hit children in prams (Lee, 2026), and one Reddit user said the gates "make you feel like livestock" (as cited in Lee, 2026). Woolworths also made a profit of $1.60 billion in the 2026 financial year, before significant items (Mickleboro, 2026). So I wanted to know whether a company that size needs to treat every shopper as a possible thief, or whether this is an overreaction.

Woolworths is not alone. Since 2023 Coles has used checkout gates that stay locked until staff check any unscanned items, in more than 100 stores by early 2024 (Bass, 2024), and Kmart and Bunnings are testing their own (Harpur & Stafford, 2025).

This report tests the case for the gates with police data. If retail theft is high, still rising and found everywhere, tighter security is easy to justify. If it is falling, or found mostly in a few places, a gate at every door is harder to justify. The data cannot show whether gates work, only where and how much theft police record.

Victoria's Crime Statistics Agency (CSA) counts every retail theft that police record. In the year ending June 2026, that was 33,006 incidents, about 90 a day. The five charts below show how fast it grew, whether it is still growing, and where it happens, down to the suburbs where I live.

The CSA only counts crimes recorded by police. If retailers report more theft, or police respond to more of it, the count goes up even when theft does not. The rates in this report are per 100,000 residents, so councils with few residents and many visiting shoppers, such as the City of Melbourne, get high rates. I call Victoria's local government areas (LGAs) councils in the text. The charts use the official term, LGA.

1. Retail theft is up 67% on 2019, but it is not the only theft that rose

Figure 1

Recorded theft offences in Victoria by type, indexed to the year ending June 2019

How to use it. Hover for exact counts. Use the buttons to switch between growth (index) and raw incident counts. Click a legend item to hide or show a line.

Chart description: a line chart of the seven CSA theft subgroups in Victoria for the years ending June 2017 to 2026, indexed so that 2019 = 100. Retail theft is the thick orange line. It falls to 67 in 2022 during the lockdowns, climbs to 184 in 2025 and ends at 167. Motor vehicle theft, a grey line, follows almost the same path and ends slightly higher, at 172. Steal from a motor vehicle ends at 141 and other theft at 130. Theft of a bicycle (101) and receiving or handling stolen goods (99) are back near their 2019 level. Fare evasion, with only 12 incidents in 2026, ends at 19. The counts view shows that steal from a motor vehicle is the largest subgroup, with 71,041 incidents in 2026 against 33,006 for retail theft.

Note. Theft offences are the CSA subdivision B40 Theft, split into its offence subgroups. Index = incidents in a year ÷ incidents in the year ending June 2019 × 100. The shaded band marks the years ending June 2020 to 2022, which contained Victoria's COVID-19 lockdowns. Data from Crime Statistics Agency (2026).

During the lockdown years retail theft fell to 67% of its 2019 level (in 2022). It then climbed to a peak of 36,313 incidents in 2025 and fell 9% in 2026. Even after that fall, it is +67% on 2019.

Retail theft did not rise alone. Motor vehicle theft rose slightly more, +72% on 2019 against +67% for retail theft. All other theft combined is +36%. Retail theft's share of all recorded theft rose from 14% in 2019 to 17% in 2026.

This is the strongest evidence for the gates. Retail theft is well above where it was before the pandemic. But it peaked in 2025, and retailers began trialling the new gates in late 2025 and 2026 (Harpur & Stafford, 2025; Rowley, 2026), after the peak had passed.

2. The highest rates are in places that draw shoppers from elsewhere

Figure 2

Retail theft rate per 100,000 residents in the 20 highest-rate Victorian LGAs, 2019 and 2026

How to use it. Switch between the rate per 100,000 residents and the raw number of incidents to see how much each council's position depends on how many people live there. Hover for exact values. Click a police region in the legend to hide or show its councils.

Chart description: a dot plot of the 20 LGAs with the highest retail theft rates in 2026. Each row joins a hollow circle for 2019 to a filled circle for 2026, coloured by police region, and every one of the 20 rose. Melbourne is highest at 1,267 per 100,000 residents (1,059 in 2019). Yarra is just behind at 1,262 (up from 714), then Latrobe at 1,065 and Stonnington, shown in bold, at 944. Port Phillip made the biggest jump, from 268 to 842. Warrnambool barely moved, from 513 to 535. A dashed line marks the Victorian rate of 459. In the count view, Melbourne (2,478 incidents) and Greater Geelong (1,678) have the longest rows.

Note. Hollow circles show 2019; filled circles show 2026, coloured by Victoria Police region. Rates use resident population, so LGAs with major shopping centres or a CBD (and few residents) have inflated rates. The dashed line is the statewide 2026 rate. Data from Crime Statistics Agency (2026).

Retail theft is not spread evenly. Ten of Victoria's 79 councils account for 42% of incidents. The highest rates are in Melbourne, Yarra, Latrobe and Stonnington, and Melbourne's is 2.8 times the statewide rate. These are places where many shoppers come from outside the council, such as the CBD, inner-city shopping strips and regional centres, so a rate per resident makes them look worse than they are. The opposite happens in big outer councils. Casey, Wyndham, Whittlesea, Monash and Knox are all among the 15 councils with the most incidents (Figure 4), but none of them makes the top 20 by rate. Casey records the third-most retail theft in the state, yet its rate of 368 is below Victoria's.

In Stonnington, where I live, the rate went from 684 to 944 per 100,000 residents, about 2.1 times the statewide rate.

For security, this means the risk differs from store to store. A gate and a full-day guard fit a store in the CBD or a big regional centre better than a quiet suburban store.

3. Where is it high, and where is it growing?

Figure 3

Retail theft rate (2026) against change in incidents since 2019, by LGA

How to use it. Type a council name in the search box to highlight it, filter by police region, or use the slider to show only councils of a certain size. Click a bubble to highlight it, and double-click the chart to reset.

Chart description: a bubble chart of the 59 LGAs that recorded at least 20 retail thefts in 2019. The horizontal axis is the 2026 rate per 100,000 residents on a log scale, running from 118 in Surf Coast to 1,267 in Melbourne. The vertical axis is the change in incidents since 2019, and bubble size is the 2026 count. Almost every bubble sits above zero. Only Benalla (-36%) and Brimbank (-10%) fell. Port Phillip is highest, at +222% with a rate of 842. Melbourne, the largest bubble, sits at the far right but is up only 38%, below the statewide +67% marked by the dotted line. Stonnington is nearby at +40%.

Note. Each bubble is a local government area (LGA); bubble area shows the number of incidents in 2026. The chart leaves out LGAs with fewer than 20 incidents in 2019, because one or two extra thefts in a small count move the percentage change by 10 points or more. The dotted line shows the statewide change (+67%) and the dashed line shows the statewide rate. The horizontal axis is logarithmic so that the few high-rate LGAs do not push the rest into one corner. Data from Crime Statistics Agency (2026).

Figure 2 showed where theft is concentrated. This chart asks whether those same places are also where it is growing fastest. On the median, they are. The quarter of councils with the highest rates grew by +79% since 2019, compared with +64% for the rest. The biggest gains in the high group came in councils that started low. Port Phillip went from 300 incidents to 967, Campaspe from 88 to 274 and Central Goldfields from 25 to 80. Two of the highest-rate councils grew more slowly than the state. Melbourne, which records more retail theft than any other council, is up 38%, and Stonnington is up 40%, against +67% for Victoria.

So retail theft is not becoming more concentrated in the worst-hit councils. The ten councils with the most retail theft recorded 43% of it in 2019 and 42% in 2026. The rise is spread across the state. Concentration is still a reason to target security, but it is not increasing.

4. The rise started at the same time almost everywhere

Figure 4

Retail theft by year in the 15 LGAs with the most incidents in 2026

How to use it. The chart opens on growth since 2019. Switch to the rate per 100,000 residents to compare levels. Hover over any cell for exact values.

Chart description: a heatmap with the 15 LGAs that recorded the most retail theft in 2026 as rows and the years 2017 to 2026 as columns. It opens on growth, with 2019 = 100. Blue cells are below the 2019 level and orange cells above it. The 2021 and 2022 columns are mostly blue. Melbourne's 2022 cell, at 47, is the lowest on the chart, and Monash is the only row that stays above 100 through the lockdown years. From 2024 nearly every cell is orange. Most rows are darkest in 2025, but Casey, Port Phillip, Wyndham and Yarra are darkest in 2026, and Port Phillip's 2026 cell, at 322, is the highest. The rate view puts Melbourne and Yarra at the top and Casey at the bottom.

Note. Rows are the 15 LGAs with the most retail theft incidents in 2026, sorted by 2026 rate (highest at top). The rate view uses a colour-blind-safe viridis scale. The growth view uses a diverging blue-to-orange scale centred on 100 (no change), so orange cells are above the 2019 level and blue cells below it. Data from Crime Statistics Agency (2026).

Figure 4 shows when the rise happened. Read the growth view left to right. All but one of these 15 councils dipped below their 2019 level during the lockdown years. The exception is Monash, whose lowest lockdown year was still 105. All 15 are now above their 2019 level. 11 of them peaked in 2025, the same year as the state as a whole, and fell in 2026.

The other four did not. Casey, Port Phillip, Wyndham and Yarra all recorded their highest count in 2026, and Port Phillip is now at 322, more than three times its 2019 level. So the statewide fall in 2026 is not happening everywhere, and the councils still climbing include the CBD's neighbours and two big outer growth areas.

When councils as different as the CBD and outer suburbs rise in the same years, a local cause such as one new shopping centre or one group of offenders is unlikely. The timing points to something statewide. The CSA figures give no clue what that was.

5. The suburbs behind the council numbers

Figure 5

Suburbs with the most recorded retail theft, 2019 and 2026, by selected LGA

How to use it. Pick a council from the dropdown. It opens on Stonnington. Hover for exact counts.

Chart description: horizontal grouped bars for the eight Stonnington suburbs with the most retail theft in 2026, grey for 2019 and orange for 2026. Malvern East is far ahead with 464 incidents (375 in 2019), 43% of the council's total. Next come Prahran with 195 (159) and South Yarra with 107 (102). Toorak shows the biggest jump, from 18 to 96, followed by Malvern (41 to 91) and Windsor (32 to 71). Armadale has 65 and Glen Iris 2. The dropdown switches to 11 other councils with high counts, and the title updates to give each council's top suburb and its share.

Note. Shows the 8 suburbs with the most recorded retail theft in 2026 within each council. The dropdown lists the 12 LGAs with the most incidents in 2026, starting with Stonnington. Each suburb is where police recorded the offence, which is usually where the shop is. Data from Crime Statistics Agency (2026).

Within a council, retail theft is concentrated in a few suburbs. In half of the 55 councils with at least 50 incidents, the top two suburbs record 72% or more of it. In Stonnington, Malvern East, which includes Chadstone, Australia's largest shopping centre, records 43% on its own. Toorak grew fastest, from 18 incidents in 2019 to 96 in 2026. South Yarra, where I live, barely moved, from 102 to 107. Even a high-theft council is not high everywhere. Most of its suburbs record far less theft than the council total suggests.

So, are the gates an overreaction?

Partly. Theft rose, but it is concentrated in a few places, so the data makes a stronger case for targeted security than for a gate at every door. Recorded retail theft in Victoria is +67% on 2019, even after a 9% fall from its 2025 peak, and 69 of 79 councils recorded more of it than in 2019. That is a real problem for the people who work in shops. The rise is spread across the state rather than concentrated in the worst-hit councils, and the statewide count was already falling when the new gates arrived.

Police data says nothing about whether the gates reduce theft, or how much theft costs Woolworths. Woolworths' profit shows it can afford gates and guards at every store. Whether every store needs them is a separate question, and the theft data points to the busiest ones. The police figures are also likely an undercount. Much of what goes missing at my store is only found at stocktake, long after the theft, so there is no single incident to report. An older survey of small Australian retailers found that retailers reported only one in five attempted shoplifting incidents to police (Taylor, 2002). What the data does show is that what we see on night fill and at stocktake is real.

What this means

If Woolworths wants shoppers to accept gates and guards, it should publish the numbers behind them, such as how much each store loses to theft and how often staff are being threatened. Until then, the police data supports stronger security at the stores where theft is highest. On its own, it does not make the case for the same measures at every store.

Methods

  • All charts use one data source, the Crime Statistics Agency (2026) workbook Data tables: Criminal incidents by LGA, year ending June 2026 (Tables 01, 02 and 03). The code above downloads it directly from the CSA site. The CSA publishes it under a CC BY 4.0 licence.
  • Retail theft is the CSA offence subgroup B43 Steal from a retail store (subdivision B40 Theft). "Other theft" is every other subgroup in B40, including fare evasion.
  • Years are financial years ending 30 June. For example, 2026 means 1 July 2025 to 30 June 2026.
  • Rates per 100,000 residents use population estimated from the CSA's own Table 01 (total incidents ÷ rate × 100,000), so they match the CSA's denominators. Rounding in the published rates changes these estimates by less than 0.1%.
  • Every comparison uses 2019 as the baseline. It is the last full year before the COVID-19 lockdowns, so growth is not measured from the lockdown low. I do not calculate a percentage change for the two councils with no incidents in 2019, Hindmarsh and West Wimmera.
  • I count LGA-years missing from Table 02 as zero, because the CSA tables leave out combinations with no recorded incidents.
  • Table 01 also has rows for "Justice Institutions and Immigration Facilities", "Unincorporated Vic" and the state total. I remove them. Tables 02 and 03 cover the 79 LGAs only, so the statewide figures in this report are the sum of those 79 LGAs.
  • Recorded crime reflects reporting and policing as well as offending, and resident-based rates overstate shopping destinations.
  • I did all processing and charting in R (R Core Team, 2024) with plotly (Sievert, 2020) and crosstalk (Cheng & Sievert, 2023).
  • The figures about Woolworths and the gates come from news reports (Lee, 2026; Mickleboro, 2026; Rowley, 2026). They give context for the question, and none of them appears in a chart.

Design

The report follows the set-up, supporting facts, main insight and solution structure in Chapter 2 of the course textbook (Mannon, 2018, as cited in Baglin, 2026, s. 2.7). The opening is the set-up, Figures 1 to 5 are the supporting facts, the verdict box is the main insight and "What this means" is the solution. It uses Segel and Heer's (2010) hybrid approach. Figures 1, 2 and 4 make one point each through their titles, annotation and highlighting. Figures 3 and 5 let readers look up their own council.

The Figure 1 heading says retail theft is not the only theft that rose, because motor vehicle theft rose slightly more (+72% against +67%). A heading that singled out retail theft would overstate the case for the gates. The Figure 4 growth scale runs from 0 to the highest index in the chart, with white at 100, so the legend shows only values the index can take and "no change" reads as neutral.

Some choices are trade-offs. A rate per resident is the fair way to compare councils of different sizes, but it inflates the CBD and big shopping centres, where many shoppers do not live. Figure 2 keeps the rate as its default and adds a button to switch to raw counts, following the warning about rates and counts in Baglin (2026, s. 4.4). Figure 3 uses a log scale because the CBD rate is several times higher than most councils. On a linear scale nearly every council would sit in one corner. The axis title says "log scale" so it does not mislead readers (Baglin, 2026, s. 4.8). Bubble area, not radius, shows counts, as s. 4.4 recommends. Figure 4 shows 15 councils, the most rows that stay readable at its height, and enough to include the big outer councils alongside the CBD. Figure 5 shows eight suburbs per council so the names stay legible.

The controls follow the custom controls and highlighting sections of Chapter 8 (Baglin, 2026, ss. 8.3.6 and 8.3.7). In Yi et al.'s (2007) terms, the buttons reconfigure each view, the Figure 3 filters filter and its search box selects, and the Figure 5 dropdown lets readers explore. Colours use the colour-blind-safe Okabe-Ito (Okabe & Ito, 2008) and viridis palettes, with line dashes as well as colour. Every chart has a source line under it, as in Baglin (2026, s. 1.9.6).

Acknowledgements

I used Anthropic's Claude (Opus 5.5) (Anthropic, 2026) to help debug R code errors, locate news reports about Woolworths security gates, check the report against the course marking rubric, and review spelling and grammar. Claude also provided technical guidance on implementing interactive features, including interactive graphs and sidebar navigation. I researched the relevant design principles using the course textbook and other references, and made the final decisions about the report's design and implementation. I reviewed, ran and tested the code and independently checked the accuracy of the sources and suggestions. The observations from my shifts, research question, analysis and final conclusions are my own. I take responsibility for the accuracy of the submitted work.

This report also draws on Baglin (2026) for its data visualisation and design approach.

Example prompts included: "Give me a list of news articles that mention Woolworths security gates" and "How can I improve this assignment's presentation, visuals and interactivity?"

References

Anthropic. (2026). Claude (Opus 5.5) [Large language model]. https://claude.ai

Baglin, J. (2026). Data visualisation: From theory to practice. https://data-visualisation.stem.melbourne

Bass, C. (2024, January 3). Coles customers raise security gate fears after getting 'locked in'. Yahoo News Australia. https://au.news.yahoo.com/coles-customers-raise-security-gate-fears-after-getting-locked-in-040349005.html

Cheng, J., & Sievert, C. (2023). crosstalk: Inter-widget interactivity for HTML widgets (R package version 1.2.1) [Computer software]. https://CRAN.R-project.org/package=crosstalk

Crime Statistics Agency. (2026). Data tables: Criminal incidents by LGA, year ending June 2026 [Data set]. https://files.crimestatistics.vic.gov.au/2026-09/Data_Tables_LGA_Criminal_Incidents_Year_Ending_June_2026.xlsx

Harpur, P., & Stafford, L. (2025, December 17). How stores fighting thieves risk putting off shoppers with disabilities and kids. Inside Retail Australia. https://insideretail.com.au/business/how-stores-fighting-thieves-risk-putting-off-shoppers-with-disabilities-and-kids-202512

Lee, I. (2026, March 5). Woolworths shoppers concerned new anti-theft gates may trap them and hit their children. The Guardian. https://www.theguardian.com/business/2026/mar/05/woolworths-new-security-gates-customer-concerns/

Mickleboro, J. (2026, August 26). Woolworths Group FY26 earnings: Sales, profit and dividend all grow. The Motley Fool Australia. https://www.fool.com.au/2026/08/26/woolworths-group-fy26-earnings-sales-profit-and-dividend-all-grow/

Okabe, M., & Ito, K. (2008). Color universal design (CUD): How to make figures and presentations that are friendly to colorblind people. https://jfly.uni-koeln.de/color/

R Core Team. (2024). R: A language and environment for statistical computing [Computer software]. R Foundation for Statistical Computing. https://www.R-project.org/

Rowley, J. (2026, June 24). Are these new Woolies gates causing more harm than good? Region Canberra. https://region.com.au/are-these-new-woolies-gates-causing-more-harm-than-good/976525/

Segel, E., & Heer, J. (2010). Narrative visualization: Telling stories with data. IEEE Transactions on Visualization and Computer Graphics, 16(6), 1139–1148. https://doi.org/10.1109/TVCG.2010.179

Sievert, C. (2020). Interactive web-based data visualization with R, plotly, and shiny. Chapman and Hall/CRC. https://plotly-r.com

Taylor, N. (2002). Reporting of crime against small retail businesses (Trends & Issues in Crime and Criminal Justice No. 242). Australian Institute of Criminology. https://www.aic.gov.au/publications/tandi/tandi242

Yi, J. S., Kang, Y. A., Stasko, J., & Jacko, J. A. (2007). Toward a deeper understanding of the role of interaction in information visualization. IEEE Transactions on Visualization and Computer Graphics, 13(6), 1224–1231. https://doi.org/10.1109/TVCG.2007.70515