Since getting into the blogging thing over on the Arcadia Project blog, Huw has been coming up with some really thought provoking posts… In Context and meaning in search, he writes:
Libraries have thousands of users and millions of resources crying out to be introduced to each other. But our search mechanisms tend to be context-less. … For users of university libraries, discovery happens in context-heavy environments such as reading lists, citations, seminars, lectures. By the time they get to our interfaces they know exactly what they want. Then they find it (or don’t).
Can we start to build some context into our own systems? And what kind of context would be useful? We can say straight off that for students the most useful context is what course they’re doing … . If we also have access to course materials (i.e. reading lists) we can really start to provide useful context for searches. How about if we have access to the content of books and articles – in particular the citations they contain? Could we start to put our searches in the context of a scholarly network based on citation?
Being a lazy sort, this brought back to mind a complementary approach I’ve discussed before, specifically the extent to which we might be able to use ad server technology to bring additional, relevant content into a page as if it was an advert (Contextual Content Server, Courtesy of Google?).
So let’s have a think again about how this might work, for example in the serving of contextual adverts on the course catalogue part of the OU website. And let’s just treat the OU course catalogue as just another What do we know?
Let’s suppose suppose that we don’t necessarily know anything, except that the prospect(?!;-) arrived at the OU website from a Google search. (I think this used to be the default assumption; nowadays, I think the OU is actually getting a lot of traffic from Facebook…)
Suppose also that we’re running a Google ad block to help target the delivery of our adverts using the Google Ad Manager (which I think is now more strictly Doubleclick for Publishers (DFP); for a primer on ad delivery, see e.g. Ad Inventory Management For Bloggers: A Comprehensive Guide). Rather than making use of AdSense to provide the ad content, suppose we’re just using Ad Manager to manage the delivery of “directly sold” ads, and furthermore that those directly sold ads are our own (OU) ads. That is, we’re going to use Ad Manager to manage the placement of local OU ads onto OU pages based on whatever information we can glean from the visitor:
Now you’ve probably heard that Google is capable of targeting ads far more specifically than just based on geography (and I’m not sure how that’s defined in the above case – is it the location of the ISP, or the browser, for example, or does the second trump the former-as-default?) or browser type, so it’s probably worth digressing for a moment here to look at Google’s ad empire.
You can basically think of Google’s ad empire as being built around three components – AdWords, where people by adverts; AdSense, a publisher service, which is the technology whereby Google places relevant adverts onto a website, and AdManager, which helps publishers manage advert placement, inventory and direct sale ads; (from what I can tell, AdSense offers a certain amount of ad management capability, but not as much as Ad Manager).
At the moment, it seems that (in the general case) Adsense demographic targeting is: a) quite limited, and b) restricted to the US:
The demographic website data used by AdWords comes from comScore Media Metrix, an Internet audience measurement provider. At this time, AdWords has demographic information on users from the United States only. For this reason, demographic site selection is available only for campaigns which target users in the United States. If your campaign doesn’t target the United States as a location, you will not see the demographic option on the Placement Tool.
However, for sites within the Google Display Network that do offer demographic ad placement, there is also the possibility for demographic targeting:
AdSense is also capable of placing ads based on content category (based on a classification of website pages depending on their content) and destination (so you can target your ads to particular websites if they run AdSense) as well as on users’ interests. Although still a crude signal, users’ interests is a technique for categorising a user based on the websites they have previously visited, and the content contained on those pages. This evolution of ad models shows how ads have moved from being targeted to particular websites, to being targeted to relevant web pages, to being targeted to the interests of the user (maybe in the context of the content of the current web page they are on). Note that where ads are targeted at the user, the website publisher has arguably lost even more control over the sort of ads being placed to viewers on their site than when the content based are being served to it (to paraphrase an oft-quoted myth(?) about supermarket analytics – a website like Mothercare could end up serving beer ads to any males aged 18-35 visiting it!;-)
So how does this help us placing
“direct sale” ads contextual content on our web pages?
For a start, using Google ad manager, we can customise the delivery of content based on locale, browser type, connection speed, or operating system. For certain computing courses, this may help us target particular ads. In the main, however, it would seem that there is not a lot of demographic information we can draw on to place content appropriately to prospective students. (That is not to say that the OU doesn’t have intelligence that we could use to target the placement of place ads if that demographic information was available. For example, we know the age, gender, level of PEQs, and possibly income level of students who have previously started OU courses, as well as their completion rates and course grades. So if we knew the same information for prospects on the course website, we could tune course and content ads shown to them on that basis (e.g. don’t show Masters courses content to 18 year olds…; do show MBA ads to 40-50 year olds from London with a prior degree and income in excess of £35k; and so on…))
Returning now to the Google Ad Manager, there is actually a way to place ads based on additional information, if we have it, by using custom targeting. For example:
You can define custom criteria based on common characteristics of your site visitors that you collect through your website. DFP Small Business can’t determine these criteria, but if, for example, you have user registration data from your site, you can specify demographic information such as gender, age and user interests.
So for site visitors who have set an OU cookie, and where we know something about them, such as their age, or gender, we can add information to visited page that will allow Google Ad Manager to target
an ad a piece of contextual content more specifically for us.
Now let’s turn to the library catalogue… if we can add a piece of information from a student record, such as an affinity string for a particular course or demographic group, to a page, we could presumably use that information as a custom criterion to support the contextual delivery of content via Google Ad Manager?
From my own previous dabblings with “library analytics”, it was quite noticeable how different audience types were identifiable, e.g. based on their referrer sites or the domain the request has arrived from. (For example, the OU Library receives a well segmented amount of traffic from the College of Law as part of a partnership agreement; and traffic from the OU campus at Walton Hall or the OU’s regional offices is identifiable from IP ranges.) Once again, this information could be used to tune the delivery of contextual content using Ad Manager. (In the Cambridge University case, if separate colleges of departments have separate IP ranges, this could be used to signal a request for contextual content delivery.)
What about OERs and something like the OpenLearn site? Can any of the Google advertising apps help us deliver more contextual content there? A new technique released by Google earlier this year known as remarketing may provide some clues as the the future here. Remarketing is an extension of the interest based advertising approach whereby a publisher signals something about a visitor to their website, and then AdSense uses this information to allow the publisher to target particular ads at those visitors when then visit another site in the Google content network. So for example, if you come to my website and look at a particular course on OpenLearn, I can signal that so that when you go to another website, I can place an advert that advertises a course similar to the one you looked at on OpenLearn…
As described here, there are lots of other ways you can approach remarketing:
– Add all site visitors
– Add users to a category
– Add only users who do not convert
– Add visitors who abandoned shopping baskets
– Target people who convert with up-sell/cross-sell messages
– Target users a month after they completed a purchase
… and then serve targeted ads to them on third party sites… but that does, of course, require spending ad pounds via AdSense… (For more examples about remarketing campaigns, see e.g. Google AdWords Remarketing Campaigns: See how we set up our own campaigns.)
One final approach for delivering contextual content that it increasingly possible for Higher Education institutions is to deliver that content within the context of Google Mail using Gmail contextual gadgets (an increasing number of HEIs are starting to adopt the Google Apps for Education suite, which includes provision of GMail to students). Contextual gadgets work by placing content based on content extracting filters as described in Gmail Contextual Gadgets Developer’s Guide. These filters can extract signals such as:
– the email address in the From: line of the email;
– the Subject: line of the message;
– the first 1,000 characters of the body of the message;
– any HTTP and HTTPS links in the subject and body of the message;
– the text “Hello World” (case insensitive) in the subject and body of the message;
and also run a regular expression over the extracted terms in order to further tune the contextual content placement.
So what’s the way forward? In the short term, it might be interesting to explore the possibility of using Google Ad Manager to place contextual content into a web page based on the IP address of of a visitor. Choosing what content to place might be informed by analysing currently collected Google Analytics, segmenting visitors by IP range, and looking for any obvious clues about what sorts of things those visitors are searching for. By looking at click-thru rates on contextually placed content/links we’d get some sort of signal back as to how relevant those links were.
Alternatively, we might consider the use of contextual Gmail gadgets, such as a gadget that detects the presence of a course code or course related URL in a message, and then place course related content into the gadget on that basis? As the contextual gadget spec includes support for Open Social and OAuth, it would presumably also be possible to serve personalised content to the user based on who they are… so for example, it may be possible to identify from an email that a user has received an email form the library, and in the widget display their library profile; or if they have received a library recall notice, complete the gadget with a form that allows for the renewal of the corresponding items?! Gulp…
PS I wonder: how does Google go about selecting what ads to run as public service ads through AdSense, and would it make sense to support a “non-commercial” AdSense programme, where e.g. publishers could opt to run not comeercial AdSense ads, but free, contextually relevant (or interest based) OER “Ads”?
5 thoughts on “Contextual Content Delivery on Higher Ed Websites Using Ad Servers”
Thanks for the mention, Tony – interesting stuff! I think you’re right that we would need to address not just what we can find out about a user, but also what bits are relevant where. I’ve been meaning to take a closer look at Google analytics … I’m using it on various things, but mainly just to monitor use rather than type of use. I have a feeling there is a pretty sharp divide between Cambridge use and non-Cambridge use for a start. And also between those who find our stuff vs those who have it bookmarked (though I use bookmarks less and less particularly on chrome)
Comments are closed.