The Rise Of Contextual Marketing

In this post we examine the techniques used to know you for purposes of improving targeted advertising.

In Pete Mortensen’s post The Future Of Technology Isnt Mobile Its Contextual, he highlights the shift in computing towards a paradigm called Contextual-Awareness Computing and outlines the four ‘graphs’ that are required before contextual computing will work, these four graphs are Social, Personal, Interest, and Behaviour. In essence, Contextual-Awareness Computing meaning computers are able to proactively react to external stimuli as opposed to being commanded by its user – these four graphs being considered the necessary information to provide relevant context. The concept lends itself well to the notion of Just-In Time Interaction/Information as touched on in Frog’s Creative Director, Scott Jenson, blog post Mobile Apps Must Die, where interactions and information are not dependent on direct input but rather your current activity i.e. your current context.

This insight is valuable to marketers and a service (albeit vastly evolved) they’ve been providing for many years, known as customer insight. The difference between then and now is that it was satisfactory to generalize your audience, main reason being that the channel the media was being published to was static and needed to be served to a large audience. In contrast, in todays world the channels we use to communicate with our customers are dynamic, dispersed, and real-time. This brings new challenges and opportunities – one challenge being knowing the ‘context’ the user is currently in, where previously it was safe to assume they were at home sitting comfortably on their couch, the opportunity is being able to ‘connect’ with your audience i.e. knowing them to provide relevant and personalised messages and offers.


Knowing more about you gives means to more effective advertising – an attractive opportunity for marketers and one reason why we are seeing so many players are entering into the market of contextual-computing, currently Google+ being the most dominate.

Lets begin by looking at the data used to learn more about the user and their current context using. We can group this data between Soft and Hard sensing, as done so in the Intel Labs post Context Awareness Activity Recognition. The following list (taken from Intels post) shows inputs for each category:

Inputs for hard sensing include:

  • Accelerometer (measuring motion)

  • Location

  • Ambient audio

Inputs for soft sensing include:

  • Device activity

  • Social networking actions

  • Calendar data

Hard sensing is data that can be obtained from the sensors on the devices, they are normally low level data streams that need to be analysed and abstracted to provide meaningful knowledge for the system. For example; sampling the devices accelerometer can determine the users current state in motion i.e. sitting, walking, jogging, running – this is derived by matching trained signatures (patterns), an example of a dataset is from Wireless Sensor Data Mining Lab, it provides samples of when user(s) performing different activities and can be used to train your model to derive what the user is currently doing.

Soft sensing is extracting information held or accessible by the device and normally in a form of information, i.e. higher level of abstract compared to hard sensing data. Some examples of this include: Browsing history, installed applications, calendar, friend connections on Facebook and your contacts, etc. This information can be used to not only derive context but also provide insight into your user i.e. interests, friends, habits. A simple example; imagine scanning through the installed applications and discovering the applications RunKeeper, Nike Training Club, and Fooducate. Based on this information it would be reasonable to assume your user is fairly healthy and possibly not the best candidate for a Dunkin Donuts coupon.

You might be asking yourself if this information is really accessible and how you go about obtaining it. The answers are Yes, and luckily other people have done most of the hard-work for you; Some examples are:

  • Qualcomms Gimbal; An SDK that can derive the users interest by examining their browsing history and context based on location.

  • Funf, now owned by Google, that offers a similar framework for capturing context.

  • Google have even integrated Activity Recognition into their Google Play Service.


The current landscape looks something like this: Imagine the following scenario – you have just wrapped up at a meeting around 13:15 and are scheduled to be in another meeting at 15:45 and it takes you 15 minutes to get there. You enjoy fresh food and regularly eat at Yo Sushi and Subway.

You leave the meeting, as you head outside you grab your phone, launch the map and start heading towards the next place. You recently updated the Subway application that is Geo-fenced’ enabled – as you approach a subway you are notified that you are entitled to a free drink with your next order. You check to see if you have enough time and decide to head in to grab a bite. Sound great right? There are a couple of problems with this approach:

  1. You are required to have the application running

  2. There is a good chance that the application has a very basic understanding of you and your context (distance and time) which dangers on sending coupons and messages at inconvenient and irrelevant times i.e. what if you didn’t have enough time.

  3. You would need to have individual applications running for each restaurant.

Of course there are alternatives such as network based location detection (and notifications) and coupon applications, but neither provide adequate intelligence (in terms of understanding the user and their context) or creative freedom that some brands require – also this is one instance of contextual-advertising.

It’s 2013 – there has to be a smarter way right?!?


Let’s imagine the same scenario but with a smarter alternative (keep in mind that all this is technically feasible right now). Now instead of individual applications running your device ‘knows’ you have time for lunch, it broadcasts interest in available restaurants on route to your next meeting (or searches). All restaurants that satisfy your criteria respond back with available offers, your device further filters the offers and alerts you of the most relevant ones. You can either browse further or accept one, browsing further will launch the associated web or native app, accepting an offer will update your route adding the next destination the chosen restaurants.


To some it might sound far fetched but those who have looked into or worked service discovery know the mechanics are available, the devices and infrastructure (cloud) are capable but two things need to happen (in my opinion):

  1. Standardisation needs to be established for advertisements (in the meantime we can perform the filtering with existing search technologies)

  2. The user must have completely control and visibility over their data in order to trust that its not being misused – there must be a central, trusted repository established for this (Google maybe?)