Marketers can use the knowledge available in one social domain to inform targeting decisions in another, letting the full richness of data in the social web bleed through an entire portfolio of digital campaigns.

For example, LinkedIn lets advertisers target its users according to more then 140 sector classifications.   We took LinkedIn’s sector taxonomy and applied it to the 70,000+ workplaces addressable to Facebook advertisers.  We then looked at how fans of presidential candidate and Ames Straw Poll victor Michele Bachmann index by sector, on Facebook.

So, for example, a Michele Bachmann fan on Facebook is 3.4 times more likely to work for a religious institution that an average Facebook user.  The results jive pretty closely with what you might expect from a Republican primary candidate like Bachmann.  She indexes very highly against  stalwart Republican sectors like energy, defense, and religious institutions, and she performs poorly against traditionally liberal sectors like entertainment, technology, and education.

I am incredibly bullish on the future of the display business. Online advertising is now the second largest advertising medium, behind television, and  the growth in display is outpacing that of the overall ad market. New technologies are enabling incredibly exciting campaigns, as well as making it possible for advertisers of all sizes, large and small, to benefit from what display has to offer. Right now, the display business is estimated to be around $20-25 billion. At Google, we believe it has the potential to grow to $200 billion, as ad dollars continue to follow consumers online, and as the barriers to entry such as complexity and cost diminish. I believe we are in a new golden age of advertising, powered by technology, and I, for one, can’t wait to see what happens next.

Remember that Google also thinks that 75% of online advertising will be social by 2015. Which I certainly won’t argue with – we believe that this is a huge shift that people have not realized yet, for the most part. I don’t see how it’s a $200 billion business UNLESS it’s very different from just putting the same messages up in a banner online – that means interactivity, creativity and virality are paramount in the new advertising paradigm.

As we consider data and targeting within the social media world, it is fun and instructive to look at the meta data on audience segments and make some inferences about how likely we are to be able to target a large proportion of our desired group. Using the example from our last post, if 54% of people drink coffee, but say we are only able to find 5% of them because they somehow exhibit this behavior in an addressable way (e.g. filling out a warranty card in the offline world tied to their home address as has been done for decades) we can say that our “find rate” for that audience is about 9%.

In social media, with audience-supplied information, find rates are going to be quite low. Thus targeting models and targeted marketing are going to flow towards those people on whom we or our partners have data. The ideal solution is a combination of audience-supplied and inferred data. And with that – I refer you to the latest posting on our optim.al blog about the broad category data targeting on Facebook.

Audience valuation, the trends underlying it and how they change are all key parts of what we do everyday at XA.net. When you think about audience value, you should think about a few components – which I helpfully frame as questions here:

1. How difficult is it to identify this audience, find the audience, and is targeting it even possible at all? (fundamental, not trivial questions!)
2. What is the premium to target this audience versus the audience that contains it?
3. How much response lift does targeting this subset of the larger audience provide?
4. Does the targeting method/medium place any special constraints on the message you can deliver?
5. What operational costs are involved with targeting the audience via this methodology?
6. Is it possible to measure effectiveness using controls, and/or does showing something to this audience make it impossible to properly measure effectiveness (a la Heisenberg’s uncertainty principle)

A silly example to think through some of the above might be let’s say, identifying coffee drinkers. So while it is easy to identify someone who drinks coffee (ask them!), finding them in our chosen medium (online) is far more difficult than offline. The value of this audience is also probably fairly low compared to the cost to target it  — 54% of US adults drink coffee daily and so the general value of someone who is a coffee drinker may not actually be that high. In Part II, we’ll share some ideas and thoughts here on other ways to think about audience targeting, and provide some specific examples.

optim.al also now has its own blog devoted to social media advertising issues – located at http://the.optim.al

a) Mild animation to get a user’s attention. Nothing too crazy.
b) An element for the user to select that makes the ad personal to them, for example to choose their age or state.
c) Interactivity: let people play a mini game in the ad, or somehow otherwise be engaged with the unit before they take an action
d) Clear call to action- what do you want them to do?
e) Don’t try to “tell a story”. It seems appealing to have a flash animated ad with a story to it, but these usually don’t work. Keep it simple.
f) Limit text in the ad. Common problem is trying to explain the product in the ad in detail.
g) Importantly, don’t assume a good designer = a good banner ad designer. Remember that all of the above has to happen in less than 40kb. Overcome the need to do this stuff yourself because you feel you should just because you have a designer on staff.

There are plenty of other best practices, but this is a start. Creative is a big driving force in performance and receives chronic underinvestment from advertisers.

This was originally posted on Quora here.

The idea is to compare the distribution of conversions or successes (e.g. sales of a product) as against the distribution of the online audience. So for a given segment, if I see 1.5% of the converted audience matching that segment, and 0.75% of the online audience typically being in that segment, I would have an index value of 200 ~ my converters are twice as likely to be in that group as an average online user.

Here is what a chart of this looks like for a single product, for example. The upper line is the conversion cumulative audience count by segment and below it in yellow is the segment distribution for the online audience. In this case we can see the first vertical line and box indicates that about 20% of the typical US online audience accounts for 40% of conversions — and depending upon what basis we pay for the segment data we have we could continue to buy data up to a point on the right.

We’re conducting this analysis for both cpmatic.com and optim.al clients right now to help come up with the unique audience-conversion concentration picture we can harness to drive extraordinary results!

The other thing they (smartly) realized is that the two products in the example above are different. For the most part, the xyz.com inventory they are selling directly is not the same as that which I get in the exchange. For those who find it constructive to think in simile/metaphor, you might think the correct analogy is the outlet store – - those Hugo Boss pants I might find in their outlet if I want to drive 45 minutes outside of the city vs. going to Bloomingdales and paying 200% more — but that’s not correct. While sometimes you find slightly damaged goods selling for cheap in the outlet, often it’s the exact same merchandise for sale more cheaply in out-of-the-way locations. To be sure, this certainly happens, but its more often akin to an airline metaphor (I like those):

Think of the publisher direct, guaranteed inventory as a seat on a full-price economy ticket to fly on a plane, and think of the remnant ad exchange inventory as a standby seat. In the first case you know what you’re going to get, it’s a published fare (perhaps you got a discount, perhaps not), you are first in line, eligible for and might get upgraded to first class, and you have an assigned seat. The remnant standby guy is only going to get a seat if there’s space available (sometimes there’s a lot, often not so much as I recall “you missed your flight on a sunday morning out of las vegas? what were you thinking” gave up and stayed sunday night) and will if they do may be one of the last to get on board, won’t have room for their rollaway and will have to check it etc. They get a lesser product; it might be very similar but there’s also a degree of uncertainty involved meaning it might be nothing.

In remnant display, the first impression you see might end up being the 8th ad impression the user saw. And you have no way to know. See my piece on AdExchanger about why real-time bid prices are overinflated.

Our job here at XA.net is to help the advertiser or agency place their (or their client’s) inventory across that priority/price curve in the locations that drive the most ROI for them, in a creative- and campaign-specific way. It’s a tough job, but we’re having fun doing it!

Rubicon joins our other real-time partners including Google AdX, AdMeld, AdBrite and Pubmatic as live direct RTB integrations.