Are You Rostering Wrong? Why Relying on Sales Data for Your Schedule Is a Revenue-Killing Mistake
There are ghosts that haunt the aisles and stalk the shelves of your establishment.
Feb 27
Tracking smartphones for the purpose of retail analytics and people counting is a growing space, and one that offers some huge benefits over the more traditional methods of people counting such as overhead cameras, thermal sensors and laser beams. Whilst smartphone tracking technology is amazing, there are some technical realities you should be aware of.
Smartphones can be tracked actively or passively. Here is a very quick rundown of active and passive tracking as it relates to the specific technology such as Bluetooth, WiFi, GPS and GSM.
Active smartphone tracking using GSM, 3G or 4G is illegal in most countries and is the playground of government security agencies (e.g. FBI, AFP etc.) as this method of tracking is considered a man-in-the-middle attack. All other methods of active smartphone tracking require a user opt-in as follows:
As such, actively tracking smartphones is not a viable solution for people counting and retail analytics. Why? Because in order to track a smartphone in this way, the customers must all opt-in via an app download or signing into WiFi. What percentage of your customers would take this action prior to shopping in your stores?
Passive smartphone tracking does not require a mobile app or user opt-in and only provides anonymous data, and can be done in the following ways:
Ok, so now that we know WiFi is really the only way to passively and legally track smartphones in order to provide valuable data, let's talk about WiFi tracking in more detail.
In short, ABSOLUTELY!
WiFi (and all radio based communications such as Bluetooth, 4G etc.) emits a circular signal, both from the sensor (WiFi router) and the smartphone. This means that it is essential to place the sensor in the centre of the measurement zone. If the sensor is installed to one side (for example, on the sidewall of a retail store), it will measure as much of the neighbouring store as your own.
If the sensor is not centrally installed it will measure neighbouring stores. The best place to install a sensor is in the middle of the measurement space, in the ceiling.
It is possible to use a directional antenna in order to push more signal in a particular direction, however these are costly, ugly (they are large and external to the WiFi router) and are quite ineffective in small spaces as their primary purpose is to extend the signal range over a large distance. As such, directional antennas are not typically used by WiFi analytics providers. Read more about directional antennas here.
In short, not very.
The accuracy of heat mapping via WiFi depends on two key factors:
The number of sensors required depends on the space being measured. For a square or rectangular space, the minimum number of sensors required is four (4). Each sensor needs to be placed in the far corner of the space and their signals need to overlap so that each sensor picks up the probes from phones in the space. If smartphone probes are only picked up by less than four sensors, the quality of this data diminishes quickly.
Whilst three sensors can also provide triangulation, the best accuracy you can expect from three sensors is to place a smartphone in a 5m x 5m box, and even that is not consistent. It is common when using three sensors to have most of the devices placed in a straight line between two of the sensors, thereby rendering this data useless.
No. Attempting to heat map a space, or even to quantify the number of smartphones within a small zone of a retail store (e.g. changing rooms) is impossible with two WiFi sensors (routers). This is simply because two sensors are not able to triangulate the smartphone. Where only two sensors are used, the smartphone may appear to be placed anywhere in the signal strength intersection of the two sensors as shown by the dotted line in the image below.
Using two sensors as shown above does not work. A smartphone seen by both sensors may be in the store next door (you have no way of knowing). Additionally, this method will not provide an accurate count of smartphones within a store because neither sensor is centrally placed.
Ok, so let's assume you've got 4 WiFi routers installed to provide heat maps, what accuracy can you expect?
Where a smartphone is connected to the WiFi network, the data quality is quite high because the phone is constantly communicating with the WiFi router and updating its location. However, where the smartphone is not connected to the WiFi network, it is only provide location updates each time it sends a probe in search of WiFi, which is every 6 to 90 seconds.
In an average retail store, with daily visitors of 50 to 1,500 people and location updates every 6-90 seconds, the results are useful, but not highly accurate. Customers appear to 'teleport' around the store and it is impossible to provide path analysis or heat maps with a very high level of accuracy.
If highly accurate (ie. 1-2m2) heat maps are important and you need accurate data to inform store layout, then you need to use camera or thermal-based people counting technologies. These technologies will provide accurate, reliable and usable heat mapping data.
Not really. WiFi can measure a range of up to 80m (depending on the antenna), however the measurement range cannot be restricted to less than five (5) meters from the access point or sensor. This is due to the dB signal strength readings from smartphones being on a logarithmic measure. As such, low RSSIs (dB) that are close to the sensor with high signal strength tend to 'blend' together or jump around so that any measures inside 5m become meaningless.
This technical reality of WiFi signal strengths make it impossible to provide accurate data for very small spaces such as change rooms.
Since 2017, both Apple and Android have progressively increased the randomisation of MAC addresses in WiFi probes. This means that every time a phone probes for a WiFi connection, it presents a different MAC address.
As a result, it is increasingly difficult for a WiFi analytics platform to accurately identify the actual number of smartphones it is 'seeing'. For example, if a phone sends 100 probes and only 10 have the real MAC address, does the remaining 90 probes represent 1 phone, or 90 phones?
In short, MAC randomisation has had a large impact on WiFi-based smartphone tracking, particularly when it comes to data accuracy.
Over the last two years, Blix has invested millions on developing our own proprietary algorithms to uniquely identify smartphones regardless of their MAC address. Blix hardware and firmware is optimised to collect the full probe packets from smartphones and the entire contents of these files are put through our machine learning algorithms to 'fingerprint' each smartphone.
Our algorithms are trained on extensive datasets across thousands of Blix sensors and utilise multi-dimensional cluster and probabilistic analysis to fingerprint smartphones regardless of whether their MAC address is randomised or not.
This means that Blix is leading the way when it comes to WiFi-based smartphone tracking and is able to provide great people counting and retail analytics data in spite of MAC randomisation.
If you are considering a WiFi-based people counting and retail analytics platform, get in touch with Blix today. We are happy to share our knowledge of how WiFi works, what is, and isn't possible, so that you invest in the best technology for your business.
