Michael MacMillan is from Vizualize which is a company that is looking at analytics. Founded in 2003, the company now operates in 40 countries – and has large customers including Marks & Spencer and Boots in the UK, for example. The company supports thousands of locations and analyses many, many individuals.
He is hearing that what customers want is not ‘omnichannel’ – the consumer wants the retailer just to know them. Doing this on the web is relatively easy, but online is not the same as offline. MacMillan showed a chart from IBM that set out how the use of analytics will develop (below). We’re probably at the point of actionable insights, he thinks.
Early attempts to use cameras for analytics failed because things took too long. Early on, the firm found that people in Germany move at 1.5m/s and so with analytics taking four seconds, the subject was 6m away – so out of range of the system. That started the company looking for viewers further out from the signage.
Traffic measurement is a key metric and retailers want to understand how many people pass by a location, the demographics of its customers but also want to know dwell times and also whether the visitor is a repeat customer. BT & Wifi tracking are relatively easy to impliment, but only a small percentage of consumers will actually be carrying enabled and active devices, so wifi or beacon tracking is not enough on its own and can vary a lot. It may range from 20% to 50% on an hourly basis. Video is the next option, but privacy concerns can be an issue, especially in Europe. Vizualize claims it can get 98% accuracy in identification but there are questions such as how to exclude staff – if you track staff, you may get caught up in industrial relations issues.
Wifi tracking is relatively easy to do, but not very precise – the location accuracy is normally 3m to 10m (not 1m to 3m as sometimes claimed, MacMillan said).
The sheer amount of data is daunting. MacMillan said that Youtube sees uploads of 100 hrs per minute, but he said that Walmart’s CCTV uploads 2000 hours of video per minute and that the total video captured in US retail is around 2 million hours per minute!! Target, the US retailer, as an example has 150,000 cameras on its own. Globally, there are 50 million cameras in retail and that means really big data if you can analyse it. Integrators of analytics cameras are looking at digital signage as a business opportunity as they already have relationships with retailers.
“What about staff engagement?” is a question that Vizualize often gets from retail clients who need to know if their staff are reacting to shopper activity?
Initially, facial recognition accuracy was just around 60% or so, but now this is up to 98% and the company has tested its software at one entrance that sees 2 million visitors per day (an entry to a Hong Kong Station). The MTR is now using the data to track passenger movements and testing against video recording has shown 98% accuracy. Chinese buyers often buy in groups with one shopper, but five people in the group, so often when you need to analyse, you need to do it fast. The analytics software can track data in real time. Vizualize has developed software that can ‘rub peope out’ of the video, leaving simply a silhouette so that users images are not recorded. All children under 14 are automatically rubbed out using height and shape.
Using this data, a user’s path through the store can be tracked. The system can analyse the video much faster than real time. This means that there is no need to store video and that can overcome privacy concerns.
The system can track engagement where a consumer is by a table (for example in a display of phones) or even by item on a table and this data can be sold to brands by stores so that the brands can work on improving engagement. The software can also track head attitude and see where the buyer is looking.
There are lots of metrics that can be tracked, from mood to age group, gender, height, and weight but also clothing. Clothing retailers are using the analytics to track not only the styles adopted by those coming into the store, but also to track fashion trends.
You need neural networks to recognise individuals and this technology is under development, although privacy concerns are likely to be a constraint on the use of this. Vizualize can make a fast count of numbers and can track traffic direction in real time. You can use a mobile camera, (think GoPro) and can then process 10 hours playback in 1 minute, there’s no need to watch the video.
As well as people, the firm does vehicle counting. Number plate recognition technology is easily available and MacMillan said that it’s now seen as a very basic technology with freeware available on the internet for anyone with a camera to be able to do it.
The company also tracks Google streetview and can look at a route from google using live dash cams. The company can analyse a distance as long as the Tour de France in 2.5 minutes, with software looking at the impact of brands and digital signage over the route!