As a bricks and mortar retailer, it's crucial that you have accurate retail data to inform your strategic decisions. Without an accurate picture of in-store activity, there's no way you can make the right decisions for your business.

In our previous blog we looked at some of the technologies available for counting customers and discussed the strengths and weaknesses of those approaches. Now we delve a little deeper into smartphone tracking, looking at the technical background to how it works, and explaining the reasons why video analytics beats smartphone wifi tracking in retail every time.

So How Does Smartphone Tracking Work Then?

There are a few different ways you might track your customers' activity via their smartphone devices, broadly broken into two groups: active tracking, and passive tracking.

Active Smartphone Tracking

  • Cellular Technology/Signals: It is technically possible to track a smartphone by eavesdropping on its mobile signals (3G, 4G, voice, SMS, and so on) using a device that essentially pretends to be a cell tower (performing what is known as a "Man in the Middle" attack). But there are two reasons why this is not an option for retailers: for one, the equipment used to do so is very expensive, but more importantly, actively intercepting mobile signals is highly illegal in Australia and most other developed countries. Unless you are the AFP and have a warrant, you won't be actively monitoring anyone's device via the mobile network.
  • Bluetooth: Bluetooth beacons have the potential to provide really quite accurate data about a device's location in your store, but the main drawback here is that in order to collect this data your customers would need to have downloaded your app and opted in by enabling the relevant permissions on their devices. Bluetooth can therefore only ever capture that very small subset of your customers who would bother to go through those steps, and it tells you nothing at all about potential customers who have never interacted with your brand before.
  • Wifi: We'll look at passive wifi tracking below, but when we talk about active wifi tracking, that means tracking devices that have actually connected to your in-store wifi network. There are a number of companies who provide wifi solutions to retail for this purpose. On the face of it, this seems like a win-win offering: the customer gets access to a free guest wifi network, and the retailer gets a whole lot of tracking data. And while technically this solution can provide very accurate location data, this approach suffers from similar drawbacks to Bluetooth. In an age of ubiquitous and plentiful mobile data, with most modern plans in Australia offering tens of gigabytes every month at very low cost, how many customers would bother to connect to your wifi network? Any data you can collect this way is likely to be for a small number of customers, and a very specific demographic subset at that.

Passive Smartphone Tracking

  • Cellular Technology/Signals: While it is technically possible to passively track cellular signals, the technology required to do this is just as prohibitively expensive as that for active tracking, and the data captured would be of limited use anyway due to encryption. Add in the fact that this approach is legally questionable and all up this is really not a viable option for retailers.
  • Bluetooth: Bluetooth beacons can potentially passively detect nearby devices (in other words, devices that haven't specifically installed your app and consented to being tracked) by sniffing for Bluetooth messages being transmitted nearby. However, this requires non-standard Bluetooth hardware and is likely to deliver poor data quality, as smartphones typically only communicate over Bluetooth with previously paired devices, so there's no guarantee that a given device would be picked up in store at all. Attempting to track devices passively over Bluetooth can only ever hope to capture a very small subset of customers that come in to the store.
  • Wifi: Passive wifi tracking is the technology used by all the major vendors of smartphone-based retail tracking systems. It is based on detecting the signals that smartphones periodically send out as they search for wifi networks. It doesn't require the device to actually connect to your network, but rather simply picks up on those wifi probes that are happening in the background.

While wifi smartphone tracking is the best of a bad bunch when it comes to tracking people in physical retail spaces via their devices, it still suffers from issues that seriously affect the accuracy of the data you will get from your retail analytics system.

Let's look at some of those issues in more detail below.

1. Sorry We Missed You?

As we discussed in our last blog, the biggest drawback with this approach is that it tracks smartphones, not people. If customers have more than one device, or don't have their phone with them, then your data might count them twice or not at all.

But worse than that, it relies on the smartphone actually sending out probes looking for available wifi networks while the customer is in the store. A typical smartphone that is being actively used might send out one of these requests every couple of seconds. If the device is locked with the screen turned off in someone's pocket or bag, or is operating in battery saving mode, then the gap can be much longer, perhaps many minutes apart, if the probes are sent at all.

Factors such as battery level, the particular mobile OS in use, as well as the specific network configuration settings on the device itself will all further impact on how frequently the device will be searching for networks and therefore whether it gets picked up by the wifi tracker.

All of which means that it is just not possible to reliably extrapolate from the data picked up by wifi sensors to an accurate count of the number of people in the store at any one time.

2. In Store? Or Next Door?

Attempting to pinpoint the exact location of a customer in your store by tracking their smartphone is fraught with difficulty.

Like all radio waves, wifi signals are circular: imagine a circle with the sensor device right in the middle. Any smartphone that it detects could be literally anywhere within that circle.

Put the sensor in the wrong place and you will be picking up devices from outside or even from a neighbouring store, and you won't be able to tell the difference (wifi signals can travel through walls, after all).

With a single wifi sensor, the best you can do is to place it as close as possible to the centre of the store. That means you will pick up devices from within a circular area 360 degrees around the sensor. Unless your store is also circular, though, there is no way that a single sensor can possibly reliably cover your entire store.

Even if you can solve the problem of only picking up devices that are actually physically located in your store, that still doesn't tell you where exactly those devices are. Typically, these wifi tracking sensors will attempt to detect the device's location by using its signal strength to tell how far the device is from the sensor. But this too is fraught with difficulty and factors such as the device OS version, its battery life and so on will all impact on the accuracy of that calculation.

In addition, the sensors can't detect distance accurately at all when the device is within 5 metres or less of the sensor. That's because of the logarithmic nature of the signal strength readings they get from the smartphone: under about 5 metres and it becomes very difficult to assess the exact distance within any kind of accuracy. But remember that we are talking about a circle with a radius of 5 metres around the sensor here. That means that for smaller stores—anything under about 25 square metres—the wifi tracking approach won't be able to given any kind of location accuracy at all.

And even if the sensor can determine with any degree of accuracy that the device is a specific number of metres away, a single sensor cannot tell which direction that is. Draw another imaginary circle around the sensor at the estimated number of metres: the phone could be anywhere along the circumference of that circle.

If you want to attempt to pinpoint the device more precisely, on a heat map for example, then you're going to need many more sensors.

3. Heat Map Accuracy

So. Can smartphone wifi tracking produce accurate heat maps of your retail store? Not really.

As we discussed in the previous section, to get any kind of heat map at all, you'll definitely need multiple sensors so you can blanket the store with coverage. At least three or four at a minimum, with many more sensors likely required for a larger store. These will need to be positioned strategically around the store.

That means that the signals from each sensor need to overlap extensively so that each device gets picked up by multiple sensors at the same time. The system would then combine all of those individual distance readings to attempt to triangulate the approximate real location of the device.

While it might be possible to get a rough snapshot of in-store customers at a particular point in time, any kind of tracking of detailed dwell times by fixture or of the routes taken by customers as they move around the store is just out of the question: location updates for a particular device will only be possible when that device sends another wifi probe. As we've already discussed, that could be many minutes apart.

If accurate heat maps are important to you, then there is only one technology that can generate them reliably for retailers: video analytics.

4. That's So Random

One benefit that providers of wifi tracking solutions have spruiked for many years is the ability to re-identify customers as they return to the store. Traditionally they have been able to do that by recording the MAC address of the smartphone device. MAC addresses are unique identifiers used in computer networking, and by recording this the wifi tracking solutions have been able to recognise specific individual customers every time they come back to the store.

However, this tracking has become so ubiquitous that over the last few years the major smartphone device manufacturers have fought back to protect the privacy of their customers by introducing MAC address randomisation.

Modern smartphones no longer broadcast their true MAC address when issuing wifi probes.

Instead they simply generate new random addresses each time, making the process of reidentification much more difficult.

This not only prevents reidentification of a returning customer, but may also lead to incorrect counts of individual devices within the store, if they use different MAC addresses whenever they issue wifi probes. That means one device in your store for a period of time could be incorrectly counted as multiple unique customers.

While it may still be technically possible to reidentify devices by recording a "fingerprint" of device characteristics recorded with each probe request, this is now much more complicated and is effectively a continual arms race between the device OS developers and the wifi tracking companies, requiring constant technical changes to the sensor algorithms as each new mobile OS release brings further changes that are intended to protect user privacy.

For A Clearer Picture, You Need Video Analytics...

Ultimately if you want to count people, not smartphones, then there is only one solution that can give you accurate data: video analytics.

WingArc Retail Analytics using the existing security camera feeds to monitor in-store activity, counting footfall and generating heatmaps that show an accurate and detailed breakdown of everything that is happening in the store.

Interested? Let's talk

Looking for Accurate Retail Data?

WingArc Retail Analytics captures accurate data on in-store activity to help you measure your marketing, reduce costs, and boost sales.

Matt Armstrong

View posts by Matt Armstrong
With over two decades' experience in the technology industry, Matt is WingArc Australia's manager of marketing and communications.

Leave a Reply

Your email address will not be published. Required fields are marked *

Scroll to top