WHAT EXACTLY IS ONLINE BEHAVIOURAL TARGETING?

It’s not surprising that many people are still confused about what is, and what isn’t, behavioural targeting. Some companies use the expression ‘behavioural’ to describe anything from the time of day people access a website (daytime behaviour) to the amount of times they visit them (heavy/light usage behaviour). With behavioural targeting being a highly popular online advertising technique, it isn’t surprising that some organisations are hoping some of the behavioural shine will rub off on them - but, in my view, it’s misleading and confuses everyone even further.

Essentially, behavioural targeting is taken to mean the ability to use previous web activity to automatically adapt website content (editorial or advertising) to make it more relevant for individual users. More specifically, behavioural advertising is about showing more relevant ads to users after they have viewed specific content or clicked on certain ads. As Crimtan is a behavioural advertising company, we will focus on that aspect here.

But wait – it’s not that simple! For purists, only behavioural advertising using pre-defined user profiles developed over time and held in a database counts as ‘behavioural’. When users are ‘retargeted’ with ads after just one exposure to a web page or ad it isn’t strictly ‘behavioural’ as no profiles of user interest or purchase intent are developed. However, as retargeting involves using cookies to deliver relevant ads I am sympathetic to the view that it should be considered as a part of behavioural targeting.

One other thing to note is that ‘proper’ behavioural profiles can also contain inferred demographic or location data which isn’t strictly ‘behavioural’ but allows advertisers to further refine targeting to specific groups of individuals.

So, in an attempt to de-mystify this confusing topic, here then is our cut out ‘n’ keep guide to behavioural advertising. If you can come up with any better definitions, or think we have missed anything, we would love to read your comments. To help interested web users even further, we have added a glossary at the end to explain the more technical terms. And in case any advertisers want to know where Crimtan fits in – our technology is capable of delivering campaigns using any of the techniques described here.

RETARGETING

This is the ability to show specific ads to people who have visited a certain website, seen a certain ad or searched specific terms online. It can take three forms

1. Website Retargeting.
An advertiser puts a ‘tag’ on one or more pages of their website. When a user visits that page, a cookie with an anonymous ID is sent to the users web browser and stored there. This cookie can only be seen and read by the website or a technology partner (often an ad network) that sent the cookie. Whenever a user visits a web page, the servers delivering the page content (including ads) look for cookies in the users browser. If the cookie and anonymous ID are present in the browser the user can be shown a specific, relevant ad. For example, if the user doesn’t click on an ad, a different ad could be shown next time. If the user does click on an ad and goes on to purchase the product advertised, subsequent ads could contain a message for a complementary product. If they didn’t purchase after clicking on the ad they could be sent a different ad - perhaps an offer. Alternatively, as it is possible to identify uses who clicked the ad and then purchased the product, it is possible to exclude them from future campaigns.

2. Ad Retargeting
This is similar to Website Retargeting except an advertisement carries a tag identifying a type of campaign or specific creative (say Car or Nokia-Apps). When a user visits a page containing this advertisement, a cookie with an anonymous ID is sent to the users web browser and stored there. The advertiser can then retarget users who have seen the ad in the same way as ad retargeting.

3. Search Retargeting
When a web user arrives at a site via a search engine after having typed a specific keyword (or phrase), an anonymous cookie ID is associated with that specific search query. Advertisers can then retarget these users with products and services relating to the search query.

AUDIENCE PROFILE TARGETING

This is a much more advanced form of targeting whereby web user interests, demographics, location or indications of purchase intent are stored as anonymous profiles in databases so that advertisers can select specific pre-defined groups of similar users and target them with advertisements that will be relevant to them. This can be done in two ways.

1. Website Profiling
Web pages carry a tag identifying certain characteristics of the page content (say travel, male or HD TV). When a user visits a website he receives a cookie with an anonymous ID (as with website retargeting). As the user browses the website, each page’s individual characteristics are stored against their anonymous cookie ID in a profile database. Each cookie ID eventually has a variety of characteristics stored next to it, so that advertisers can buy users grouped together by pre-defined characteristics.

2. Network Profiling
This is similar to Website Profiling, except a number of different websites are collected together to form a network of websites. Just like Website Profiling, pages are then individually tagged according to their content and, when any website in this network is visited, an anonymous cookie ID is sent to the user. With Network Profiling, more users receive a cookie with an anonymous ID, more characteristics are gathered more quickly and relevant advertising can be shown on any website the user visits across the whole network of websites.

Search terms can also be used in Website and Network Profiling to enhance user profiles. If web users arrive at a site (or sites) via a search engine, characteristics suggested by the search term can be added to the user’s anonymous ID profile.

ADDITIONAL DATA LAYERS

Certain companies specialise in supplying additional anonymous user data, gathered both online and offline(such as age group, gender, location and other known interests), which can be added to existing web user data to enhance anonymous user profiles.

While all the types of targeting described here can be conducted anonymously using only non-personal data, it is possible to combine this anonymous data with data known about an individual – for example if they have registered for a site and given personal details. By law, a web user has to give their prior consent for a company to use their personal data for Behavioural Advertising (or any other service).

GLOSSARY

Tag: A small piece of code, usually JavaScript, contained on a web page. This code effectively asks the server for a cookie or associates an existing cookie with a specific characteristic.
Cookie: A small file of letters and numbers downloaded on to your computer’s web browser when you access certain websites. Cookies allow a website to recognise a user’s computer. A cookie itself does collect information.
Web Browser: The software you use to access the internet and display websites. Common browsers are Internet Explorer, Firefox and Chrome.
Anonymous ID: The string of numbers or letters stored on your browser’s cookie that identifies your PC. As only the website that issues the ID can read an ID It is anonymous as you cannot be identified unless you also give that website your personal details.
Server: The computer hardware and software that stores and delivers (serves) content, such as web pages, over the Internet. The content of a web page can be supplied by a number of different servers in different locations and is assembled to create a finished web page in your browser.
Ad network: A company that connects a number of websites and enables advertisers to advertise across them using only one point of contact.
Characteristics: Indicators within page content that point to a users lifestyle interests (e.g. eco-friendly, travelling, DIY, cooking) or intent to purchase something (e.g. a hire car, a DVD, a camera). Page content can also indicate specific age groups or whether the user is likely to be male or female – allowing an inferred demographic to be added to the profile
Non-personal data: Any information that does not enable the user to be personally identified.
Personal data: Information you have supplied that, either by itself or in conjunction with other information, can uniquely identify an individual. (e.g. name, address, email address or telephone number).

DATA, DATA, EVERYWHERE...

If you are all at sea about DSP’s and are drowning in a pool of new data then this piece is for you. A few years ago it was said that the people with the most data would rule the world. Now everyone in online advertising has heaps of web user data and if you want even more it’s not difficult to get. It’s got to the stage where we are all using the same data - we’re just slicing it up in different ways. But is more data better, or is some data better than others and what exactly do all these ‘data companies’ actually do?

Who feeds the DSPs?
The ease at which anonymous user data can be gathered and traded online has given rise to any number of suppliers who have been buying data, organising it and selling it on to a variety of eager buyers in agencies, ad networks and ad exchanges. These data aggregators accumulate data from a number of online and offline sources (such as surveys, retail records, lead generation companies, travel booking websites) and package them up into groups according to lifestyle, demographic, industry, company size and so on. They then sell this to anyone across the entire spectrum of ad inventory managers – from DSP companies to ad networks, exchanges and yield optimisers. Some of the major players in this business include Nielsen, Bluekai, Exelate and Experian.

This recent proliferation of data suppliers has given rise to more and more companies who use this data to enhance inventory quality and offer it to advertisers. They are loosely termed Demand Side Platforms (DSPs) and they add a serving and optimization layer on top of aggregated inventory and data sources. Advertisers can access this data via desktop interfaces that enable web audiences to be bought conveniently and efficiently. DSPs can be roughly divided up into these two groups – but some companies perform both functions.

Audience trading and optimisation companies. Enable agencies to buy specific inventory from a number of sources (including across ad networks and exchanges) to create their own custom networks for advertisers and optimise that data during a campaign. Companies include B3, Cadreon and Vivaki.

Bid optimisers.
Allow advertisers to bid for unsold or undervalued inventory on an impression-by-impression basis, across ad exchanges. The inventory sellers get guarantees that the impressions will be sold for the highest bid price above the specified reserve. Companies include appnexus, Dataxu, MediaMath and invitemedia.

While the emergence of DSPs is great for advertisers in many ways it does mean that almost everyone is using the same pool of users and any (anonymous) information associated with them. While rates might be low, so is the inventory quality and the agency has to train staff to manage the systems and optimisation – and take the media risk relating to placement, delivery discrepancies and click fraud. This potentially offsets the efficiencies that attracts advertisers to a DSP solution in the first place. And, of course, DSPs are not free.

What now for adnets?
It remains to be seen how readily UK agencies embrace DSPs and how this affects the existing ad network landscape. Undoubtedly some of the bigger agencies will test and begin to use DSPs for some campaigns and some ad networks will suffer as a result. One thing that won’t change is advertisers thirst for unique, premium inventory combined with great segmenting and optimising technology plus, of course competitive rates, detailed campaign reporting and great customer service.

These attributes remain in the hands of ad networks with exclusive inventory deals and the latest tech – ad networks like Crimtan.

BEHAVIOURAL ADVERTISING, PRIVACY AND COOKIES.

Why is everyone STILL talking about Online Behavioural Advertising (OBA) and privacy? And why exactly are cookies back in the spotlight again? The technology is complicated and, because it’s difficult to understand how it works, some people worry what it's getting up to. Also, OBA has become caught up in some of the concerns about web privacy generally – particularly those surrounding the proliferating use of social networks. OBA will play an increasing part in online advertising (as a recent Forrester report confirms - see here) and all of us who work in the industry should feel confident enough to be able to talk about and, if necessary, defend the use of Behavioural Advertising. So we thought it was a good time to take a step back and address the subject objectively.

Firstly, the purpose of OBA is to deliver web users ads that are more relevant than most of the ones they already get. Hopefully, this means that web users have a better web experience while advertisers waste less money and publishers get paid more for their ad inventory. So what’s the risk? How is it possible to do this without somehow ’spying’ on what web users are doing online? Now this is where it can get complicated - and different OBA companies use different technologies - but there are two fundamental things to understand. Web users are anonymous and no personal data is involved – unless you give it to the company yourself.

Personal data and anonymity.

Let’s take a look at these two issues separately; first the anonymous part. If web users are anonymous how can we ‘identify’ individuals to assign lifestyle interests and judge possible purchase intent as they browse the web and then deliver them specific ads rather than random ones? This is where cookies come in, as every time a web user visits a site in the OBA company’s network, they are sent a cookie that includes a unique number. A cookie can only be read by the site that sent it and it can’t collect any data from the user’s PC.

Every time a user then visits another website in the OBA company’s network the ID is recognised and more interests can be added and existing interests examined to see if there is a relevant ad that can be delivered. This means a web browser can be ‘identified’ but not an individual. However, there are two further issues that concern some people. Can you be identified from your interests and are flash cookies somehow ‘worse’ than browser cookies?

This brings us to the ‘personal data’ bit. Some OBA companies are also websites or online service providers (like Yahoo!, Google and AOL) so if you have registered with them and gave personal details (like your name and address) they may associate this personal information with your cookie ID and your interests or search terms. Everyone has heard about the AOL case where a New York Times journalist traced an AOL user from their published search queries. This was only possible because AOL stored actual search terms and these were specific enough to enable the individual to be identified. But other OBA companies don’t ask for or store any personal data to start with and Crimtan, like many other OBA companies, just stores inferred interests. Site URLs and search terms simply indicate an interest, and it is this interest that is stored alongside the cookie ID – never the URLs or search terms themselves.

But what about Flash cookies?!

Flash cookies and Behavioural Advertising are getting a lot of attention at the moment, as some people have suggested that they somehow have a greater ‘impact on consumer privacy’ than regular cookies. Of course, Flash cookies behave exactly the same way as browser cookies – they are just stored on the PC rather than the browser and can hold more information if need be. So there are no greater privacy implications than with browser cookies. Despite this, Members of the GPP for OBA (see next paragraph) have agreed not to use Flash cookies to help collect data for Behavioural Advertising.

Reassuring concerned users.

So what is the industry doing about reassuring people who are still nervous about Behavioural Advertising - and what else could we be doing? Well it’s important to remember that even web users who have no idea about Behavioural Advertising and do nothing are at no risk whatsoever. The worst that can happen is that they start getting useful advertising not rubbish! However, for whatever reason, there are some web users (a tiny minority) who actively oppose the idea of Behavioural Advertising, cookies and even advertising – and it is reasonable that they should be able to opt-out if they wish to. That’s why the IAB has established the Good Practice Principles for Online Behavioural Advertising (GPP for OBA – see here for more) and member companies are audited to ensure they comply.

The key Principles cover Transparency, Notice and Choice. Transparency means that the company has to declare on their own site and partner’s sites that they are involved in Behavioural Advertising. Notice means that users should be clearly informed about what data is collected, how it is collected and how it is used. Choice means that users are given the opportunity to opt-out via a simple link. In an ideal world this should be something you only have to do once but, unfortunately, most companies use an opt-out cookie that needs to be replaced (by repeating the opt-out process) if a web user deletes all their cookies.

Respecting consumer choice.
Crimtan has led the industry by implementing a unique opt-out that respects user choice in the long term by ensuring they remain opted-out even when they delete their browser cookies – a first among Behavioural Advertising companies.

For users who aren’t prepared to trust that they have actually have been opted-out by the Behavioural Advertising company they can, of course, opt themselves out by altering their PCs browser settings to reject third party cookies. Any users who are also concerned about Flash cookies can go here to delete and reject these as well.

Summary.
So there you have it - in a nutshell, Behavioural Advertising:

• doesn’t use personal data ( unless you gave it to a company and agreed they could use it).
• is totally anonymous for users who have not offered their personal data.
• has a set of Good Practice Principles that leading practitioners of Behavioural Advertising comply with.
• can easily be opted out of – permanently in Crimtan’s case.

Anyone worried about privacy online could spend their time more profitably examining the privacy policies of social networking sites and free webmail suppliers, or anywhere else that has your personal information – both online and offline.

We're interested to hear the views of other people in the industry. Are people wrong to be concerned about Behavioural Advertising – or should the industry be even more pro-active in reassuring web users? Let's hear your view.