S1 Ep 6
Video Analytics in Retail - A Look at Current and Future Use
In this podcast Colin Peacock and Professor Adrian Beck consider how the use of video analytics is broadening across the retail estate, especially in grocery at the self checkouts.
They discuss some of the feedback and insights from 70 retailers from across the globe and consider how the new technology and analytics are increasingly being used for loss-prevention, security-management and even measuring queue length.
With the growth of AI and other new technologies, they discuss how well-planned video analytics systems can really support retailers, as long as their objectives are well-defined at the start.
Podcast summary
Video analytics in retail has been “the next hottest thing” for years, and this episode shares what retailers say they’re actually doing. This survey of 70 respondents gives us a holistic view of current best practice and video analytics in retail benchmarks, including where next.
Colin Peacock and Professor Adrian Beck discuss where video analytics in retail is most used today (especially grocery self-checkout), how non-grocery differs (more distribution centre and perimeter use), and what retailers are planning next.
They also cover expanding use cases like people counting and queue monitoring, and why improving tech quality and shifting mindsets matters.
Cleaned transcript
Why video analytics in retail keeps returning
COLIN:
In this episode, I’m joined by Professor Adrian Beck, and we discuss the latest findings from our survey of retailers looking at the use of video analytics in retail.
How are you?
ADRIAN:
Very well, Colin. Good to see you again.
COLIN:
Good to see you.
We had our video analytics benchmark survey results shared by your good self last week. Again, lots of interest in this topic, video analytics.
I think we’ve been talking about video analytics for at least 15 years. Heat maps, perimeter controls, and video analytics is going to control your store.
We were told 15 years ago, I think just after we were told that RFID was going to be bigger than the Internet or bigger than the computer or something.
So it’s had quite a lot of publicity, the whole world of video analytics.
This survey that you completed got responses from 70-odd retailers and how they were using or looking to use video analytics. The report will be published in a few weeks or so.
What were your high-level takeaways from the responses to the survey and the activity we’re seeing?
What the video analytics in retail benchmark survey covered
ADRIAN:
We had responses from 70 retailers, so a reasonable sample, but not quite as many as last time.
Again, we were able to split them between grocers versus non-grocers. That was a good split between them.
Video analytics is a very hot topic at the moment, isn’t it? People are talking a lot about AI and machine learning and everything will be solved by the use of artificial intelligence, apparently.
Video analytics overlaps with that in some ways. They’re similar, but they’re not the same.
This survey was really interested in trying to understand around the entire retail estate how, and to what extent, retailers are using a wide range of video analytics.
We asked them to comment on 47 different types of analytic covering six different places, from the distribution centre through to the front of the store, in-store checkouts, in the back of the store, and so on.
So we’ve got a real flavour of the extent to which people are currently using these analytics, and those that are planning or trialling them in the near future, so some intention as to what’s going to happen with these.
Where video analytics in retail is used today: grocery vs non-grocery
ADRIAN:
Perhaps not surprisingly, in grocery the checkout is still very dominant in terms of the space where most are trying out a range of video analytics, particularly related to their self-checkout.
How can you try and identify non-scanning, mis-scanning, walk-aways, and so on? That was perhaps not unexpected.
For the non-grocers it was much more their use around within their distribution centres and the surroundings around it. Perimeter control, unauthorised doors being open.
So it’s much more about the fabric of their sites rather than within the store around self-checkout.
That was mirrored as well in terms of what people are planning and trialling out.
We saw even in non-grocery people beginning to look particularly at self-checkout, but also interestingly in-aisle use of video analytics.
There’s this growing trend, I think, for people to be interested in seeing whether they can use the analytic to identify suspicious behaviour in the aisle and then bring that to the attention of members of staff who can perhaps go and intervene, offering lots of customer service to try and put people off.
COLIN:
You and I have both, we went to London, didn’t we, to the Amazon Go store, the Tesco store, put Easter eggs, more little Easter eggs in our pockets and walked out and still got charged for them.
So I guess there must be some trickle-down technology that might work, as you say.
I thought it was also quite interesting. There was a couple of examples, additional use cases, perhaps, that weren’t in the 47 around people and maybe the more progressive use of people counting.
But also we had some examples of how they’re looking at the self-checkout area to understand what the self-checkout hosts and whether the self-checkout hosts were there or not. I thought that was quite interesting.
Improving video analytics in retail: from binary alerts to complex aisles
ADRIAN:
Traditionally these sorts of analytics have been very security focused. It’s been predominantly around trying to deal with safety and security issues.
But increasingly I think we’re seeing the use of analytics for non-security. How can the retail organisation benefit from the information that these things can deliver? People counting, like observing when queues may be getting too long, and you can then begin to get staff into those spaces.
It’s still predominantly around safety and security, there’s no question about that. But I think a number of retailers are now thinking about how this data, once you’ve put the network in, once you’ve invested in the equipment, it’s then a question saying to what extent can we use this for a range of purposes.
What struck me listening to the responses from those on the call was that the technology is improving, the quality is improving.
One of the challenges is that the ability for the analytic to work effectively is really compromised when it’s put in quite complicated spaces and has to do really quite complicated things.
If you think about protecting a building, you either want to know whether somebody’s there or not. It’s quite a binary bit of an analytic, and they’re very good at doing that.
But when you bring it into the aisle and you start saying to the system, please tell me when somebody’s been a shoplifter, that’s very complicated. Lots of things go on in the aisle.
Often shoppers will look like shoplifters, particularly if you’re offering something like mobile self-checkout, for instance, which can make it look like people are stealing things.
I think what we’re seeing is a maturing of the quality of the technology now where it’s able to be tasked to do more complicated things in more complicated environments, and I think that’s going to be the big change moving forward.
Hence why we’re seeing people beginning to use it quite effectively in the aisle. It’s now beginning to get better at doing the things you want it to do.
If it doesn’t, it can end up as a real distraction. You spend more time trying to make it work than actually benefiting from what it’s doing.
We heard that in some of our earlier research, where people said because it’s so complicated to get it to work well, you spend an inordinate amount of time just trying to get the thing to work.
So I think what we’re seeing moving forward from this survey is the ability of the technology now to be tasked to do more complicated things in more complicated spaces, and seeing some benefits from that.
Mindset shift for loss prevention leaders
COLIN:
I think we also heard as well, a loss prevention and security mindset, if you’re the leader of video, how you need to change your mindset and how you need to change and think about how your video and your security cameras can actually serve a different purpose.
That felt like to me that was a real learning for some of the people who are working in the loss prevention space, to turn their minds: well, if I was a consumer market research person, how would I be able to think about presenting my service and my camera technology to those people?
We’re going to have a session on that in July, the whole piece around how to change your mindset as a loss prevention leader and security leader. One of the speakers has agreed to come along and speak to that.
Thank you, Adrian. We look forward to the reports and talking about our favourite subject of video analytics in the future.
ADRIAN:
Thank you very much.
COLIN:
Cheers, Colin. Bye-bye
