Last week, we discussed how Grouping is the first of four things we should consider before over-targeting without sufficient data and testing. This week we will look at the second consideration: How Referring Traffic data may trump over-targeting when considering how best to serve our advertisers.

Referring traffic – it is helpful to consider where the user was before they came to the content on which we serve them an ad, as this may reveal what they are in the right frame of mind for, or "their intention" to use online marketing-speak.

To consider the intention of the user, it is helpful to visualize that person running out to catch a pass; they are moving and we need to hit them with the ball where they are going to be. By the time that user sees our ad on that content, what is their intention? Where were they before they came to that content? What drove them to that content that day, that moment? Were they looking for entertainment? Shopping? Information? Connecting with friends? How did they discover this content? Through a link on a search engine? Through a recommendation on a social media page? Are they already a loyal follower of that content? Do we show a different creative if the person comes to that content via a tablet vs. a PC?

While we can use our analytics tool when studying our own site's traffic, marketers need comScore for directional guidance on this when looking at the traffic that comes to someone else's site…and it's even less exact when trying to determine the referring agents driving traffic to sites within networks. Considering the types of referring traffic: site visitors often come by circuitous routes along the way going through Social Media (currently #1 referring agent); Search (for some sites, paid traffic drives the most visitors, for others, clicks on organic links drives the most traffic – in both cases the users are primarily attracted to the content, not the site); another publisher site; shared media or from that user's Favorites, etc. Should we consider the effect of referring less than or more than the selection of demographic target? Maybe that user is moving in the direction we (the marketer) need even if the user is not the exact demographic profile we thought we wanted to target or the content was not exactly what we envisioned? And, of course, does serving ads on mobile devices mean that we should consider altering the demo target by device? And, if so, how do we know for sure before we run the campaign what is the right target?

Backing in the right creative to match various intentions seems intuitively easier than guessing which exact target will convert best against a single creative, even if the conversion is proportional like following a brand on Twitter. Given that, why the over-reliance on targeting rather than referring? Is there a shortage of creatives? Can't we create and test more? Maybe the landing page, which didn't convert our target would work against a slightly different target with the right referring criteria?

Is this another example of how over-targeting may be costing the advertiser in terms of lost conversions as well as lost learning opportunities?

So far, we have focused on how Grouping and Referring should be considered before tight-targeting requirements are locked in. Next time we will look at the secret sauce behind behavioral targeting – Taxonomy – and why we need to consider this before insisting on a "too-tight" BT target.