Live facial recognition in retail: Building the facial recognition business case and keeping public trust

Facial recognition in retail is one of the most debated and fast-moving areas in loss prevention today.

In this episode of the Sell More and Lose Less podcast, Colin Peacock and Professor Emmeline Taylor explore what it takes to deploy live facial recognition responsibly and sustainably, drawing on insights from an ECR Retail Loss meeting attended by around 50 retailers from across the globe.

They break the conversation into three areas: organisational readiness, the facial recognition business case, and maintaining public trust at scale.

On readiness, they discuss why deployment is not about flicking a switch. Decisions about who receives alerts, how watch lists are managed, and how sensitive data is safeguarded all vary widely between deployments.

On the facial recognition business case, the numbers are encouraging: one retailer reported a 20% reduction in shrink, and another saw losses fall to 0.7% in stores where the technology was deployed. But regulators are pushing back on some use cases, insisting that facial recognition must be proportionate to legitimate purposes such as reducing violence and improving in-store safety, not just delivering financial returns.

On sustainability, Emmeline argues that retailers must go beyond compliance and adopt a transparency-first approach. Successful deployments in markets like New Zealand have been built on clear signage and upfront communication with customers. She warns that surveillance creep, where law enforcement or other parties seek access to retailer systems, is a real risk that must be managed through clear information-sharing agreements. Drawing parallels with the early days of CCTV in retail, they argue that facial recognition will likely follow the same trajectory: controversial now, expected within a generation. But it takes only one retailer going rogue to undermine public goodwill, making minimum standards essential.

Facial recognition in retail: readiness, ROI and trust

Getting ready: organisational alignment

COLIN:

Hello and welcome to the ECR Retail Loss Sell More and Lose Less podcast series. My name is Colin Peacock. I’ve worked in the retail and FMCG business for more than 40 years, working on the dark side of retail, on retail loss at every point in the supply chain.

I’m with Professor Emmeline Taylor and we’re looking at live facial recognition: where the industry’s at around the world in terms of deployment, use, and the business case. We’re going to look at three aspects. First, how organisations are getting ready and aligning themselves to what is actually quite a big undertaking. Second, how they’re building a return on investment and business case. And finally, as they look to scale and deploy across many years, how do they sustainably maintain the trust of the public and the court of public opinion? How do we keep this going without undermining trust in retailers and their use of this technology?

Let’s start with getting ready. What are we hearing about organisational alignment?

EMMELINE:

This is a really important point, and one that I think is a road less travelled in terms of the conversations on live facial recognition. The ECR Retail Loss meeting was attended by a truly international audience. There is global interest in live facial recognition.

One thing that came through loud and clear is this is certainly not just about flicking on a switch. It can be years in the making. That organisational alignment is really key because what some of the participants talked about was how suggesting or tabling the idea of live facial recognition suddenly prompts a retrospective look at all the policies and procedures that need to be revisited. Just one case in point: the training of staff, or the safeguarding of sensitive data.

One person said we talked about there being maybe twelve different principles that retailers need to explore to have a safe, ethical, and sustainable deployment. They said each one of those principles would prompt its own working group and its own meetings. This is actually quite a big undertaking, but get it right and it could well be worth the investment in time.

No cookie-cutter approach

COLIN:

To your point, there didn’t seem to be one cookie-cutter approach. You couldn’t produce a step-by-step guide, because decisions like who do you deploy the alerts to? Store staff, managers, security guards, or remote monitoring? There are lots of complications.

EMMELINE:

The offerings are so different from the vendors. The algorithms are operating differently. Any business looking to deploy this really needs to do their homework around how it actually operates and what they need to have in place.

There were two really key points of departure. Some areas around deployment are relatively uniform, whether you’re in California or Auckland. But two areas still have a lot of diverse activity.

First, who do the alerts go to? Do they go into the store? Can they be shared with third-party security officers? Is it in a SOP? Is it remotely centralised? The performance of LFR is going to be different depending on who gets that alert. Some businesses say they get the alert just for information, just for awareness. Others get the alert and it prompts a series of interventions or interactions with a known offender. Hugely variant.

Second, watch lists. Who maintains them? Who has control over them? What warrants somebody being put on a watch list, and for how long? And how can somebody be taken off? We’re seeing a lot of divergence there as well.

COLIN:

I don’t think I’ve yet heard of a project that’s taken less than 12 months to put together. Often it can be three years. Someone’s going to spend a lot of time doing those twelve meetings. You’ve got to be quite motivated.

Building the business case

COLIN:

Let’s come to the second area. How do you build an ROI and a business case? I think that links back to some tension or dilemmas in terms of what might be seen by others as a business case versus what you might see internally.

EMMELINE:

This is really key. That business case that might land well internally might not be one that’s supported by regulators. The key principle for any deployment is around legitimate use. It has to be proportionate to the reasons why you’re using it.

We’ve seen some cases where regulators have said using it for things such as fraudulent returns simply doesn’t satisfy that legitimate use. It might be good internally, it might have a financial ROI, but regulators are saying maybe this is more about violence, about reducing aggression or enhancing in-store safety. That’s quite difficult to put an ROI on.

We’re seeing these quite murky issues around the ROI. But we did have some figures shared. One retailer said shrink was down by 20%. Another said their shrink was down to 0.7% in the stores where it had been deployed. That’s quite a good return on investment. But again, it’s balancing it with proportionality. What other types of criminality is it reducing?

COLIN:

It’s also worth thinking through the total and true cost of deploying it. It’s about cameras, of course, and not everyone’s cameras are up to spec. But it’s also about the time and effort of people involved in the governance of the data. There’s a lot of cost here. Understanding the true cost would help. Those numbers certainly were very encouraging.

Sustainable deployment and public trust

COLIN:

Let’s come to the third area. Imagine the organisation is signed up, the privacy piece is on board, the CFO is happy, the regulator is happy. How do we sustainably scale and deploy without losing the trust of the public?

EMMELINE:

In terms of sustainability, what is really key is to move beyond compliance only. This is such a fast-paced, moving area. Several nations are now scrambling to legislate around the use of LFR and other powerful AI-driven technologies. It’s about going beyond the bare minimum. Don’t use the current regulation as a checklist to work against. Think ahead. What would be a really ethical, privacy-first way of deploying this?

Key to that is transparency. Where we’ve seen successful deployments, looking towards New Zealand for example, they were so keen to have customers on board from the outset. This wasn’t about deploying it in the shadows. They were upfront. You could not ignore the signage. It’s so in your face. There’s nothing to hide.

Having that really sensible conversation up front with consumers is so key, because if you don’t, that’s when suspicion sets in. You can bust all the myths around live facial recognition by having the public on board from the outset.

We talked about having twelve principles driven from the UK Information Commissioner’s Office around surveillance cameras. Using those as a launch pad into discussions around LFR is probably a really good, sensible start to having a sustainable and responsible deployment.

Surveillance creep and law enforcement access

COLIN:

Maybe if criminals start realising this technology is there, they might be dissuaded from going into stores. But I did want to touch on something. It would seem like great technology. Why wouldn’t retailers just say the police can have access too, to see if there are missing children or criminals? Is that a concern or an opportunity?

EMMELINE:

I think it’s a real concern around function creep, or surveillance creep as it’s sometimes phrased. Businesses will really need to be clear and hold the line around who they share information with. There are some use cases where law enforcement say they’re in the store constantly because they want to use our systems. We need to be really careful about that and have clear information-sharing agreements in place.

I think it’s really important for the public to know the safeguards around who can access this very sensitive personal information. Let’s ensure it’s for really clear purposes and it doesn’t creep into other uses, as attractive as they might be to begin with.

COLIN:

There’s also a distinction between live facial recognition and retrospective facial recognition. Retailers are already being asked to share CCTV for evidence collection. Maybe there’ll be less sensitivity about retrospective, but still there are safeguards.

The CCTV parallel and the road ahead

EMMELINE:

Exciting times. I think, used responsibly and through a cynical lens of being very risk-aware of what could potentially happen and mitigating for that, this could be a really exciting space to begin to finally tackle those prolific offenders, those organised criminal networks, and those violent offenders that can often move with impunity between different locations because it’s almost impossible to track them through just human intervention. This could be really powerful.

COLIN:

My sense from the sample of 50 or so retailers: three years ago, probably one or two were thinking about doing a trial. Today, I think at least 30% in the room are actually doing it, another 40% are about to, and the rest are waiting in the wings. It does feel like certainly in the USA, UK, New Zealand, maybe soon Australia. Europe seems less well-placed at the moment. But the passion from those who had adopted it was quite profound.

EMMELINE:

We can definitely draw parallels to CCTV. If you think back to the 90s, there were huge civil liberties questions. Is this really desirable? Fast forward 30 years and the question is now: why don’t you have CCTV? Something happens, a crime takes place, and the public asks why didn’t we have cameras?

I absolutely think that will be the case for facial recognition in 30 years, in those countries that are supportive of it. But now is the time to ensure we have minimum standards and that all retailers hold the line, because it takes one retailer who goes a bit rogue to undo the goodwill from the public and from regulators. This is a real opportunity to get it right for the good of those businesses, but also for the good of communities in which it’s operating.

Outro

COLIN:

Thank you very much, Emmeline. We’ll speak to each other soon. Thank you for your time today and for being part of this work.

EMMELINE:

Thanks, Colin. Bye.

COLIN:

Well, that’s all we have time for this week. Thank you for joining us and thanks to all the ECR Retail Loss community for the sharing of their knowledge, their research, and latest learnings that ensure that we become truly a global platform for learning.

My name is Colin Peacock. This is the Sell More and Lose Less podcast. If you liked it, please subscribe so you automatically get all future episodes. And if you think your colleagues and friends might also like it, please ask them to follow ECR Retail Loss on LinkedIn and tune in to our next podcasts. If you didn’t like it, then let us know and we’ll try and fix it. Either way, see you next time. Thank you again.