Capturing accurate footfall data has long been a goal for retail landlords and developers. Property Week meets one of the founders of Hoxton Analytics, a company that believes it has cracked the problem.
The Holy Grail for retail landlords and property developers is capturing accurate footfall data. There are many different systems out there, but most have flaws, which means the data is always skewed.
One company that thinks it has created a product to address this issue is Hoxton Analytics, co-founded by Owen McCormack and Will Thomas, who studied machine learning together at UCL.
The company, which works out of the Cisco-backed tech workspace IDEALondon, has devised a way of extracting incredibly accurate footfall information by capturing images of pedestrians’ feet. Property Week catches up with McCormack to find out what’s afoot.
The problem was: how do we gather accurate, useful data on the pedestrians who visit a location, without resorting to invasive surveillance? This is even more important now with the new General Data Protection Regulation rules. Our solution was inspired by looking at Google Street View. To preserve privacy, Google blurs out the faces of pedestrians, but it’s still easy to tell some broad demographics about people - gender or clothing style, for example. We purposely avoid any faces by installing cameras low to the floor pointing downwards, and we count and profile people based on the footsteps and shoes we see go past.
The first step was to prove we could extract accurate people-counting and demographic information algorithmically from videos of the floors of certain locations. Through free customer trials, we were able to demonstrate accuracies for people counting of more than 95% and gender detection above 80%. From the strength of these results, we raised capital from a retail-specific accelerator called TrueStart, developed our minimum viable product and proved demand via early customers across property and retail.
We install small cameras about 50cm off the floor, pointing down at the ground. All the cameras see are people’s feet - walking into a shop, for example, down a street or into a shopping centre. From these images, our machine-learning and computer vision algorithms accurately count the people - in and out - and can tell certain things about them, such as gender, based on shoe style. Our customers then access the data we generate either via our online portal or our application programming interface if they require more flexibility.
The data is useful for understanding how busy a location is, what kind of people visit and how this is affected by factors such as weather, marketing activities, seasonal events or new developments in the area.
Our footfall counting is exceptionally accurate - 95%-plus - and due to a system of manual auditing and checks, it doesn’t drift over time; it actually improves. Our clients also appreciate the additional insights such as gender. One of our biggest differences is the form factor: it is very straightforward to deploy the camera on a wall, even in challenging outdoor environments where overhead counters and break-beam sensors wouldn’t work, and there are privacy or accuracy issues with surveillance cameras.
We have deployments in shopping streets as well as shopping centres, train stations and retailers. We’re mainly based in central London, with a handful of deployments outside London.
Our main clients are retail property landlords: FTSE-listed clients who use our technology to gather metrics on their shopping centres and high-street properties. We also have clients in retail and transport, such as train stations and airports.
Fantastic. There is a clear gap in the market for auditable, non-invasive insights that technologies such as mobile-phone tracking can’t provide. This is reflected in the fact that we currently have a 100% conversion rate from trial stage to contract.
We are constantly developing new insights within our framework of ‘accurate and non-invasive’. By collecting more granular profile data - eg commuter, leisure visitor, family unit - we will give our clients the data they need to test their activities. This allows the client to better understand performance. For example, they can see the success of a marketing campaign, how weather affects footfall and who are their most valuable customers. We have just scratched the surface in terms of the data this technology can generate. People flows, dwell times, location occupancy, event-based footfall and detailed customer profiles are just some of the features on the product pipeline for development over the next year or so.
11 September 2017 | Updated: 16 September 2017 10:44 am
10 April 2015
27 February 2015