Thought Leadership: “Behavioural Targeting – Describing the user with words”

Agencies are reporting an upswing interest in behavioural targeting as advertisers wrestle with the economic imperative to get even more value from their online campaigns. The push for performance right across display of brands, through to click ratios and actions, means the opportunity to target users by their interests is much more appealing than ever before.

But let us remind ourselves that the industry is still beset with the inappropriate display of significant brands against articles about murder, racism and other content – which simply is just not productive or efficient. Some networks stand accused of damaging the brands via misplacement. Behavioural targeting should be an extension of this problem or opportunity: we should place the ads where they will be valuable, not where they are less useful, or damaging. Page by page, user by user.

Google has the keyword model right – you do not choose words like murder or racism to place your text ad! Instead you choose the words that reflect the brand values and commercial product or service you are championing. You choose the criteria for relevance. Simple.

So surely we must apply keywords to display: avoid placing ads on those IASH problem-pages about murder, and go for just appropriate pages where the right words, context and relevance matches the ad campaign requirements. Should we buy ads by the website, the website section, or instead determine relevance page by page, user by user?

Just as Google leverages the statement of intent through search words, so the keywords people read on pages of content start to reveal a sense of intent and interest. It might be Google search result links that users are reading; or the pages within a burst of research, ahead of an anticipated purchase; or more likely the regular digest of certain types of news.

Fundamentally the keywords we read over time indicate a good degree of insight about who we are and what we like. Analysing page views linked to cookie IDs in real-time means technology can stay abreast of every change in a single user’s interests, as keywords can be grouped into clusters of words and hence themes of different interests, all for the same one user cookie ID. We still do not know who the user is, but we have a cloud of keywords like a halo around one.

Some behavioural technologies are plain simple. They pre-divide the world into categories and then group websites or large swathes of a website into a section like “Sports” or “Football”. People who go to certain URLs get higher “Football” scores on the histogram of categories. But there is little or no analysis of the actual words on the page. The IASH problem-pages persist.

And the bigger problem is that some “concepts” or categories appear everywhere. There are “fashion” stories in the main news as Michelle Obama wears her latest frock, or user generated content that bypasses the publisher’s own taxonomy efforts to pre-segment a website. How do we find these lost segments unless we read the keywords like a user reads them?

More fundamentally for publishers, most of the page view impressions happen on pages with no large amount of keywords. Home Pages point users to a variety of navigational options, or Section Heads provide snippets of where the deep content actually sits. So the behavioural opportunity is to re-target people on the Home Page, based on what they like reading, habitually or behaviourally, at the deeper content level: especially when article level reading might represent less than 25% page view volumes on a content rich site.

Analysing the best words on a page, and placing them as a keyword summary or tag cloud around each cookie ID, has some important opportunities. Without having to pre-categorise the universe of content to a single taxonomy, publishers and advertisers alike can start to create their own categories on the fly. If you want a James Bond category to coincide with the launch of a new film, then craft a few keywords together as the campaign brief, then use behavioural keyword technology from vendors like Grapeshot to target those pages or users who have some of those words in their Page DNA, or User DNA. Not surprisingly the Click Through Rates go up as the ad is targeted to certain pages or users, and ad wastage is reduced.

But alas, behavioural targeting often sounds like cherry picking the best ad slots by achieving relevance for the user, at the cost of not achieving the volume of ad placement. Not that appealing to publishers and advertisers who need to shift the volumes each month! Well, working with keywords, the next innovation is to treat ads, with their keyword campaigns, as competitors. Just like ads compete on a Google search, so display ads can compete against each other to be the most relevant for a given MPU slot. The keyword campaigns behind each display ad compete to have the most overlap with the words Grapeshot extracts for each Page DNA or User DNA profile.

Agencies buying their usual volumes can use the contextual keyword-based behavioural targeting to “hot-swap” different creatives and campaigns between the ad slots they have purchased – making placements within existing volumes more automatically relevant to certain pages and cookie IDs. It can be the trigger to swap in a different ad format, or a different creative: but the decision is based on relevance of page or user. Some have called this an AdSense-like functionality, but on steroids, as we can capture the behavioural patterns in real-time, and at a keyword level. Keywords behind the ads target the Page DNA and the User DNA.

If we know the keywords on a per page or per user basis, then we can do more than help DoubleClick place the right display ad: we can introduce related content to read, related video-clips to pre-roll, relevant white-papers to download, or affiliate product recommendations: the agenda becomes relevance within prescribed volume buys.

Keywords are driving one half of the market: it is time Display used keywords to describe it’s own audience. Both the publisher and the advertiser can use the URL and cookie ID of the ad slot to build out a picture of the ad’s audience: each user’s behavioural keywords and an up-to-date real-time view on what words describe their interests, immediately “now” or over the past 30 or 60 days.

Do you have the words to describe your audience? Do you have a behavioural view of each user’s interests, described in words?

Now is the economic time to put users’ keywords into Display.

Media Playground: www.media-playground.co.uk