Facial recognition trial reduces harm by 16%, justifies use – independent evaluator's report

• Findings follow six-month facial recognition (FR) trial at 25 Foodstuffs North Island stores 
• Independent evaluator says FR prevented >100 serious harm events, including assault 
• Evaluator concludes the benefits of using FR outweigh the downsides, justifying its use 
• FSNI’s latest crime stats for July-Sept show total retail crime incidents up 41% YoY 

 

Facial recognition technology reduces serious harm in supermarkets by an estimated 16% and has strong public support, justifying its use, according to independent evaluators of Foodstuffs North Island’s (FSNI) trial.  

 

FSNI’s six-month trial, conducted from February 2024, involved 25 New World and PAK’nSAVE stores throughout the North Island. The goal was to determine whether FR could effectively reduce serious harm in stores while respecting customer privacy through strict processes.  

 

The final evaluation report by independent analytics firm, Scarlatti, which monitored and evaluated the trial, found there was “strong quantitative evidence” that the use of FR reduced serious harmful behaviour in trial stores by an estimated 16%. 

 

Scarlatti’s report says a total of 1,742 FR alerts of offenders and their accomplices were generated across stores during the trial – an average of 70 a store – of which around 50% resulted in the person being approached. 

 

Scarlatti Director, Dr Adam Barker, said the 16% harm reduction was equivalent to around 100 serious harm events averted, including assaults, abuse and other aggressive or disorderly conduct. 

 

“Around half of that reduction can be attributed to actual interventions by supermarket staff, approaching someone who’s a repeat offender before they can do more harm, with the remainder due to the deterrent effect, stopping them returning,” Dr Barker said. 

 

Scarlatti noted nine instances of someone being approached but found to be the wrong person, two of which resulted in them being asked to leave, with all nine being attributable to a failure in the human process. 

 

But Dr Barker said such risks were outweighed by the benefits of using FR – justifying its use. 

 

“While our best estimate of harm reduction was 16%, a nationally representative survey of over 1,000 adults undertaken as an independent research study by OnePicture, found 79% were accepting of FR being used in retail settings even if it reduced harm by less than 1%.”  

 

Foodstuffs North Island’s General Counsel, Julian Benefield said the co-op had in recent years seen a sharp increase in the level of retail crime in stores, particularly violent and aggressive crime.  

 

“We have a moral and legal duty to do all we can to keep our teams and customers safe. The goal has been to learn if facial recognition can reduce harm while respecting everyone’s privacy,” he said. 

 

“Scarlatti’s conclusion that FR prevented over 100 serious events at just 25 stores over a six-month period shows the potential for harm reduction across our wider store network.” 

 

Scarlatti’s report said there was no detectable change in sales volumes at trial stores relative to other Foodstuffs stores, suggesting few customers changed their shopping preferences due to the technology being trialled. 

 

Benefield said the trial showed the potential to make retail stores safer, while highlighting the possibility for human error if strict rules aren’t followed. 

 

“Of those misidentifications where we approached a customer, we acknowledged our mistake and apologised as soon as possible.  We have also implemented further training and changed our verification processes as a result. 

 

“With 1,742 positive alerts, we’ve gotten it right over 98% of the time, allowing us to identify repeat offenders and intervene early before they could cause further harm in a store.” 

 

FSNI’s latest retail crime statistics, for the July to September 2024 quarter, show the total number of incidents was 4,933, up nearly 5% on the previous quarter, and up 41% on the same period in 2023. Repeat offenders were responsible for almost 38% of all incidents, including 554 cases of breaching trespass. There were 40 assaults during the quarter, up from 29 a year earlier. 

 

“Every case of assault and abuse has an impact on at least one of our team members – they can even be life-changing events. Our stores must be allowed to take the reasonable steps available to them to try to ensure that doesn’t keep happening,” Benefield said. 

 

Foodstuffs North Island will now await the findings of the Privacy Commissioner’s public inquiry into its facial recognition trial, before deciding on any further use of FR in its stores. 

 

As announced by FSNI in September 2024, the 25 stores that were part of the trial are continuing to use FR in the interim, using the same privacy protocols and processes they used during the trial. The list of stores using FR will continue to be publicly available on FSNI’s website.