S1 Ep 3
In-store Facial Recognition
In-store facial recognition technology is being rolled-out globally.
Join Colin Peacock and Professor Adrian Beck as they discuss how retailers are using in-store facial recognition for persistent offenders within the legal constraints to reduce theft, especially from persistent offenders.
The ECR working group has been tracking the use of facial recognition in retail for the last five years.
We have seen some strong results reported back to the group from retailers such as Rite Aid in USA, and then more recently updates from USA, UK and Australian retailers.
With the pressure on governments around the world to support requests from retailers for help with the growing levels of violence, abuse and theft from retail locations, there is some evidence that governments are listening more to requests from retailers to help progress facial recognition technology (click here).
Podcast summary
In-store facial recognition is increasingly being discussed as a practical tool for tackling external theft, with some UK retailers now openly sharing their experience and reporting positive results, particularly relating to facial recognition for repeat offenders.
The conversation explores the move towards shared watchlists and databases, alongside concerns about governance, data quality and proportionality.
Colin and Adrian also argue that accuracy is not the same as impact, and call for clearer, more robust KPIs to show what these deployments are really changing in stores.
They close by confirming they will revisit the topic again on 15 January 2025.
Cleaned transcript
Session recap: in-store facial recognition for repeat offenders and watchlists
COLIN:
We had a great session last week on our favourite subject we’ve been covering now for at least six years: in-store facial recognition and the use case in retail around watchlists, and aggregating cases using facial recognition to tie together all the different offences.
We had 60 people on the call. Lots of retailers, mainly retailers, and we had law enforcement, some lawyers and academics.
What were your notes, and new things you heard versus previous years?
Retailers’ attitudes shifting on in-store facial recognition
ADRIAN:
It’s an ongoing topic, isn’t it?
It’s one of those areas where we’ve watched retailers mature in their thinking around this.
If we go back three or four years, there was a lot of nervousness about organisational reputation associated with using in-store facial recognition.
We struggled to find any retailer who was prepared to go on record to say they were using it, didn’t we?
Over time, it’s beginning to become more established in some countries, and we heard from a couple of retailers in this country who were openly and quite freely describing their experience.
It was really good to hear people now on the ground using it and beginning to feed back how it works, the ways in which they’re using it, and whether they think there’s value in it.
The message from them was pretty positive. They felt it was making a real contribution to trying to deal with this growing concern we’ve got in this country and others: the amount of external crime that’s going on, and particularly those more determined, regular thieves who retailers really want to try and get out of their stores.
So their story was one of success.
It was interesting to hear how it’s becoming a little bit more established and people are prepared to accept that it could be part of their strategy in trying to deal with this external theft issue.
Shared databases for in-store facial recognition watchlists
COLIN:
It did seem as though there was a shift in a slightly different approach that I heard.
They were using a shared database as opposed to their own database.
Did you see that as new and different, or did it seem to enable the privacy lawyers of these two companies to feel better about it?
ADRIAN:
I think there was a bit of a mixed opinion about this.
Those who are saying we need to use a broader database so that, if somebody offends down the high street, they can’t just come up the high street and offend in my store, want this shared database of offenders, of people on the prohibited list as it were.
I can see why people would want to do that, because people do move around and these persistent offenders will simply go from one retailer to another.
But we heard others who were a little bit more cautious, and more concerned about how you ensure the validity of who’s going on to these lists.
You’re then relying on a third party to make sure they have good processes in place to make sure they’re not putting the wrong people on it.
This issue of proportionality came up as well, didn’t it?
If you have somebody who has stolen a bar of chocolate in your store, do you then want them to go on a list where they’re banned from every single store in the country?
That’s one of the dangers, of course, when you have these combined lists: you can become quite draconian in the way that these are used if you don’t get proportionality right.
Some retailers aired that concern, that if they were going to use these shared lists, they had to make sure they were happy, and their businesses were happy, that the quality of the information going into it was sufficient.
But some were also saying what an advantage it was, because certainly some of these organised gangs they were talking about would move around the country, and it didn’t make sense just to only have your list.
It was important to understand how other people were being impacted by these groups as well.
In-store facial recognition across countries: a football analogy
COLIN:
We heard from other countries as well, and you came up with a football analogy around the use of in-store facial recognition across the world.
Do you want to share it?
ADRIAN:
It may only make sense to people in the UK, Colin, but there seem like different leagues of willingness to use this technology.
In the UK you’ve got the Premier League, those who are really advancing, and that looks like China and the UK in particular: the ones who are moving forward with enthusiasm from the authorities.
The British police are very keen on in-store facial recognition, and increasingly a number of retailers are coming out openly saying they’re now using it.
Then you’ve got the next division down, the Championship as it were: countries who are stepping forward slowly, still wrapped up a little bit around the privacy issues, like Australia, where it’s currently on hold because it’s been reviewed by the data commissioner.
In the US it’s a very patchy picture. In some states some retailers have been banned from using it, because of the way that they’re doing it, and also like Canada, where again they’re gently moving forward cautiously, one might argue.
Then you’ve got the lower division, the non-league as it were: basically the rest of Europe, Scandinavia, where it’s just very problematic to be using this sort of technology.
We know there are huge limitations on sharing video footage anyway, and so at the moment they’re really saying it would be highly problematic for us to be using this sort of technology in our stores as things stand.
COLIN:
Yeah, it was a great session.
It really was great to hear that it does feel, slowly at least in the UK, we’re inching forwards, and the rest of the countries as per your divisions.
Measuring the impact of in-store facial recognition rollouts
ADRIAN:
One other thing I’ll pick up on is we need a bit of clarity and a bit of consistency, rigour and robustness in trying to understand what impact these technologies are having.
A lot of the people on the call would say it works, it’s successful, but actually when we began to look at some of the key performance indicators, it was less clear how they were measuring its impact.
Retailers can say, “Yeah, it’s making a huge difference.”
So I think we need to make sure that we’re getting some robustness in how we’re saying, “How is it working?” and what the measures of success are for these sorts of technologies.
People often talk about, “Our measure of success is the accuracy.” But that’s really a measure of its performance, not a measure of its impact.
So I think we need greater clarity across the industry about how we’re measuring whether these things are having an impact or not.
And I think as we go forward, hopefully that will happen.
COLIN:
Yeah, I was really confident that would happen, because the rollouts of these companies were quite significant: 200 or 300, 400 stores.
So hopefully there’ll be some stores with and some stores without, and so you’ll get some real understanding of any impact on sales, which didn’t seem to be coming up as a particular concern.
But also impact in terms of financial performance on the levels of losses.
So yeah, that was a good final point.
Next session date
COLIN:
We’re doing this again in 2025. Can you believe it? It’ll be around before you know it.
I think it’s the 15th of January in 2025.
Thank you very much for your notes, Adrian. Much appreciated, and we’ll see you next time.
Take care. Bye-bye.
ADRIAN:
Bye. Bye-bye.
