Behavioural (behavioral) online advertising

By; Dave Chaffey

In the article, we will look at the benefits available from behavioural targeting of online ads through a series of examples of how different advertisers have trialled this relatively new approach.

IntroductionI blame the first Internet ad – the feted ‘banner ad’ on Hotwired in 1994. Since then the ‘banner ad’ label has stuck for a long time. But for me, the simple label ‘banner ad’ has restricted the adoption of Internet advertising. For a start, the phenomenon of ‘banner blindness’ is well known – as users of the Internet, we filter out areas of the screen we recognise as ad rectangles and concentrate on the text – one of the reasons for the growth in Pay Per Click keyword advertising. Furthermore, ‘banner ad’ implies a static ad of fixed format rather like a traditional classified ad. Given these limitations, most media owners, digital marketing agencies and industry bodies now refer to Interactive Advertising which is more suggestive of the range of options for rich media ads, data capture ads and large format ads such as skyscrapers. This brings us to the topic of this WNIM article. Interactive advertising really starts to take advantage of the medium when the advertiser can interact and respond to the web site visitors’ behaviours and characteristics. This is, in a nutshell, is behavioural targeting; the capability to serve an ad, in real time, by recognising the characteristics of a site visitor and then serving a relevant ad.

What is behavioural targeting?Behavioural targeting is all about relevance – dynamically serving relevant content, messaging or ad which matches the interests of a site visitor according to inferences about their characteristics. These inferences are made by anonymously tracking the different types of pages visited by a site user during a single visit to a site or across multiple sessions.



Beyond tracking page visits and repeat visits to a site, behavioural tracking systems can also record events on a site such as completing a registration form or interacting with different types of content like a Flash demo.



Other aspects of the environment used by the visitor can also be determined such as their location, browser and operating system.



Contrast this with more traditional ad targeting on a site where the main method of targeting is according to the content type on a particular area of the site.



For example, traditionally, a hardware manufacturer has been able to book ads to be displayed on the IT channel of a news site, but would not be able to display ads to the same visitor when they are elsewhere on the site. Behavioural targeting enables an advertiser to target ads when that visitor moves elsewhere on the site or returns to the site, thus increasing the frequency or number of impressions served to an individual in the target market. Previous research has shown that at least 3 exposures are required to maximise impact, but often it is difficult to achieve this frequency if an ad is only served in one area of the site.



Behavioural targeting can be used either on a media site to serve ads relevant to visitors or on a destination site to deliver personalised messages about products or promotions. On a media site such as MSN (www.msn.co.uk) behavioural targeting can be used to serve ads according to pages visited in different channels such as Dating, Cars or Money. On a destination site such as Amazon, messages are displayed according to previous books browsed on the site as part of its personalisation capabilities. In this article we focus on behavioural ad targeting on media sites.



The best way to understand the principle is through an example. The FT.com example shows how behavioural targeting can be combined with other targeting based on content (editiorial), geography and demographic targeting (from the profile of customer characteristics).

FT.com launches behavioural targetingIn September 2004, FT.com upgraded its facility for targeting using a behavourial targeting solution from Revenue Sciences.



In this approach, eight segments have been identified for advertisers according to types of content visited. These are Business Education, Institutional Investor, Information Technology, Luxury and Consumer, Management, Personal Finance, Travel, Private Equity.



The new behavioural targeting facility can be combined with editorial targeting, geographic targeting, and demographic targeting.



Netimperative (www.netimperative.com) quoted Shauna Monkman, global head of online sales for FT.com, as saying: ‘The opportunity to target specific groups with the most relevant campaigns is one of the major advantages of online advertising. Not only will it enable our advertisers to target more effectively, but it will also improve the experience for our users, who will be served with more relevant advertisements.’



Companies who have signed up for behavioural campaigns to date include NTT Docomo, IBM and 02.

How well do behavioural targeted ads work?
The aim of behavioural targeting is naturally dependent on the campaign objectives, but generally, dynamic ads do improve ratings of awareness and response. Since behavioural ad targeting is relatively new, most of the effectiveness studies are based on testing built into campaigns rather than analysis across many campaigns. Here we look at two examples.



GlobeandMail.com (2004) reported on research by Dynamic Logic that suggests increased ad awareness and brand favourability.



In the US, iVillage, a portal offering content for women ran a test on behavioural targeting for Snapple meal replacement drinks. Here ads were delivered through behavioural targeting to people who had previously visited the diet and fitness section of the site, but who were reading other sections of iVillage at the time the targeted ad was served.



Behavioural ads were found to be more effective than those simply placed permanently within iVillage's diet and fitness section. According to Dynamic Logic:

76 per cent of consumers who received targeted ads displayed an awareness of them, against 66 per cent among non-targeted consumers.
Thirty-six per cent of targeted readers reported a favourable impression of the brand upon viewing the ad, against 21 per cent in the non-targeted group
Intent-to-purchase based on the ad was 37 per cent for targeted readers against 29 per cent among non-targeted consumers.


iMediaConnection (2004) highlights another example. In this case, a Dallas Mitsubishi car dealer used contextual advertising within the automotive section of DallasNews.com to deliver targeted ads. The newspaper served the same ads, when visitors returned to DallasNews.com even if they were outside the automotive area. The dealer used Tacoda's Systems' Audience Management System targeting capabilities to reach visitors whose prior actions suggested a keen interest in the advertiser's message.



In this case it was reported that the response rate among the target audience was 7.7 percent as compared to the national average of 0.33 percent. These ads played a part in doubling the number of credit applications the dealership received, and increased the number of online searches by 17 percent. The DallasNews.com campaign generated 44 percent of the total phone calls into the dealership, at a time when several other automotive promotions were running in other media.

Real Time OptimisationMany behavourial ad targeting providers describe their services as offering “Real Time optimisation”. Behavourial targeting is by definition Real Time targeting since the technologies used have to identify the viewer and serve an appropriate ad at the time of the visit.



For example, European provider Twin London (www.twinlondon.com) which uses technology licensed from Pointdexter Systems (www.pointdextersystems.com) describe two types of Real Time Optimisation.



First, media optimisation. Here a visitor is identified according to their behavioural or demographic profile and an ad is served according to the segment or profile cluster they best fit into. If the visitor does not match a profile the ad impression is passed backed to the publisher who can then publish other inventory.



Second, placement optimisation. Here the most relevant ad is served for the viewer. Poorly-performing ads can be replaced by the best performing ads and offers for each audience.



Peter Jones of Twin London describes their predictive optimisation engine (POE) as follows:



“POE makes a decision between the time a web page is loaded and an ad is served. It is a fully automated process based on pre-defined optimisation rules based on campaign marketing objectives and deliverables. What this means is that, rather than buying impressions, you are buying viewers based on response profiles, which achieves better targeting, increased response, lower CPA and increased ROI.”



Pointdexter Systems worked with a US ISP to place ads on media sites according to whether the visitor profile suggested they had potential for Narrowband, Broadband, Premium services or other membership options and the type of offer they would be most likely to respond to, e.g. anti-virus control, music content, etc.



Using this approach reduced CPA by 59% in comparison to historical data and a control group maintained throughout the campaign. Through proving the returns the ISP also increased paid media spending by 10 times and rolled the program out its internal media properties.



In a European test, MSN Spain used Real-Time Targeting from Touch Clarity. Speaking at the London E-metrics summit 2004 (www.emetrics.org), Louise Brown, portal business manager at MSN described how Real Time Optimisation service Touch Clarity (www.touchclarity.com) was used to maximise sign-up to paid service MSN8 on msn.es.



Customers profiles are built in real-time based on observed behaviour based on:

Frequency and recency of visits
Previous channel and product interest
Browser type, screen resolution, time of day and IP address.


Different creative propositions were presented to different users to maximise sign-up, for example, shared browsing, Encarta or Free Trial.



Again real time, behavioural targeting gave significantly better results. Product sales achieved through Real Time targeting increased 45% compared to the control group.

The future for Behavioural Targeting
Currently, there are many successful test campaigns being reported and these campaigns are being rolled-out more widely.



A key issue for the future of this approach is Trust. If customers realise they are being tracked across different sites (often the most powerful data) and their data sold then this may result in a privacy outcry like that faced by Doubleclick a few years ago. For this reason, some providers of behavioural targeting such as Revenue Science have placed an emphasis on trust and they warrant that no customer is tracked across sites and that the customer details will not be sold on.



Other providers such as Tacoda Systems have sought to take advantage of the more detailed understanding of customer behaviour by analysis across a network of US publisher sites and use this to target more closely. Tacoda has launched Audience Match which will give advertisers the ability to target text ads based on specific demographics. This service is similar to Google and Overture's Pay Per Click advertising, but instead of bidding for keywords, advertisers will compete for specific demographic information or audience clusters. Tacoda, of course realises the importance of Privacy and stresses that it does not collect personal information and allows concerned users to opt-out of cookies used for tracking in Audience Match.



So, we are definitely at the early adopter phase with this new technique and there is a lot of potential for modelling visitor behaviour in more sophisticated ways. Also we may see behavioural ad targeting techniques more widely applied to destination sites, once lower-cost packaged versions become available.