When the principle was first explained to me, I was running an agency display media team at a search agency that was focused on direct response clients; I was therefore interested in tactics that involved precise data points as a way to focus on user intent.

Search retargeting seemed to fit the mold perfectly: target just those individuals with display ads who have actually searched for the terms that were relevant to the client, eliminating nearly all wastage from the plan.

We were building what we called the agency’s “foundation layer” of display: site retargeting to fix on-site conversion, search retargeting to prospect and plug the leak from SEM, and social retargeting to add further scale to the audience. So we picked five clients who had a pre-agreed testing budget and rolled out search retargeting with an early vendor, only to see four out of five of the campaigns bomb!

The primary reason was that in Search Retargeting 1.0, there was no scale in the data and little effort invested in the media placement. The campaigns were great when spending $100, but as soon as the vendor tried to scale to fill the budget, they would have to broad match and lose the relevancy, and of course the ROI.

Growing The Data – Search Retargeting 2.0

But sticking with it, campaigns began to perform better over time, and in almost direct correlation with the quantity of data that was available. Now we could focus more on the relevant terms and ignore some of the broad head terms. And as any search marketer can tell you, volume comes from the broad terms, but ROI comes from the specific. With data volume no longer such a problem, search retargeting 2.0 was on the horizon.

The theory states that search retargeting should outperform most other display placements because of its accuracy, and even come close to the performance of your search marketing efforts — as an industry it simply was not there yet.

But with the intersection of search and display looking like the future of digital, I left the agency to go help make this work.

The Data High

Too many marketers went through a phase of being high on data, believing that a single reference point was all that was needed to generate great DR results. They lost sight, in their excitement, of the continuing importance of creative messaging and the context of the media placement. Knowing who to talk to is important, but doing that in the right environment and with the right story really matters.

As popularity in search retargeting grew, so did the funding, which allowed the successful players to build their own DSP (Demand Side Platform) technology to manage the quantity of data and build in these essential elements. Marrying thousands of keywords with thousands of potential ad placements is not easy though, particularly when you have to do it in real time and at the keyword level.

Advanced Optimization — Search Retargeting 3.0

In a recent Chango search retargeting campaign from a large retailer, the need for keyword level optimization is clear:

Search retargeting example 1:

• ‘clothes shop’ — CTR of 0.87%
• ‘clothes shopping’ — CTR of 0.25%

Search retargeting example 2:

• ‘shoes mens’ — 0.16%
• ‘mens shoes’ — CTR of 0.21%

The search marketer is used to a world where this type of analysis is commonplace, but what is different is the choice of media sources.

In SEM, you choose from two major engines and then can add the extended network, usually by just ticking a box and forgetting it. But with real-time display, including search retargeting, we can buy in excess of 100,000 QPS (Queries Per Second — a simple measurement of media capacity).

Therefore managing search retargeting campaigns today is complex. Typically a campaign will need to be optimized manually once a day, but then “machine learning” must be used to balance the multitude of options available.

In our examples above, the term “clothes shop” clearly had a better type of intent that “clothes shopping” for our client, but that could only be determined by analyzing the placement on tens of thousands of sites. The balance of people and technology provide the scalable solution (but interestingly also blur the line between agency and vendor).

In addition, search retargeting 3.0 leverages dynamic creative, but unlike a typical dynamic setup, there is actual search data to work with, producing richer and more relevant experiences for the end consumer. Search retargeting sprang out from a sea of providers buying on the exchanges, but now seems to be leading in terms of what can really be achieved.

A Blessing & A Curse

Search retargeting would probably not be the name our micro-industry chooses if it got to choose again. When media planners hear it they immediately get excited, as they know a good media plan should always include some type of retargeting. But their first assumption is that it targets their existing site visitors. Many conversations begin by saying this isn’t the retargeting you thought it was!

But once marketers understand, they see its value for the long term. Like site retargeting and the SEM program itself, it typically becomes an evergreen program, running continuously as a reliable source of revenue.

In Summary …

Search retargeting arrived on the media scene less than two years ago, and less than one year ago for most media planners. It leverages the power of search and executes it with the scalability of display. It is enormously complex because of the volume of both keywords and media placements, so early campaigns were often not successful (v1.0). But as the industry grew, so did the data, and with it came a certain amount of reach (v2.0).

Today, major brands invest hundreds of thousands per month on evergreen and seasonal campaigns in search retargeting thanks to the results that in-house DSP bidding technology allows for true keyword level granularity. Machine learning, dynamic creative and lots of experience means that version 3.0 is upon us … and growing.

Marketers buy keywords (for now on a CPM basis) and their ads show up at the top of Twitter searches as sponsored tweets. It’s paid search. And I can see the future SMX Advanced panel now: “Best Practices for Twitter Search Campaigns.”

Apparently Twitter has “huge” search volume, according to COO Dick Costolo, though no specific numbers were provided to AdAge. That may well be but they’re not yet huge enough to show up in comScore’s extended search rankings list:

Now for the AdSense part, which will reportedly come later.

Promoted Tweets will show up in users’ Twitter feeds (and in third-party clients) and probably be contextually and/or locally relevant. The company is going to be extremely careful about this part because of the spam and Facebook-style “user-outrage” potential. I’m sure that Twitter won’t blow it, however.

Predictably the proposed AdSense-like dimension of Promoted Tweets is causing some controversy among the punderati. But what’s striking to me is how obvious and largely uncreative these approaches are. Along those lines the prize for the best headline on this topic goes to GigaOM with “The Twitter Ad Model Revealed (What Were You Expecting, a Pony?).”

Google and Overture/GoTo before it made the world safe for paid-search advertising, which now constitutes the by far dominant component of online advertising revenues at 47 percent last year.

These business model choices that Twitter is making are both safe and proven. According to comScore figures quoted by the NY Times, “Twitter.com had 22.3 million unique visitors in March, up from 524,000 a year ago, and that does not include the millions more who use the service through third-party smartphone and Web applications like TweetDeck or Tweetie.”

Though Promoted Tweets will at first sell on a CPM basis, making more money for Twitter than CPC would, just for fun let’s do some CPC math.

Google’s revenues per search (RPS) have varied over time and have been estimated at 9.5 cents and 12 cents at different points. Google’s RPS are higher than the industry average. But lets assume 10 cents per query as a round number that Twitter might be able to get, which also assumes 100 percent fill rates.

Let’s say, using those figures and assumptions, that Twitter had 100 million searches per month (just under five searches for every Twitter user). That would translate into $10 million in CPC revenue on a monthly basis and $120 million per year. Nice numbers to be sure, but would that justify the billion-dollar valuation that the company currently has?

Postscript From Danny Sullivan: Lots for me to digest here on my first day back from a week’s vacation! I’ll probably be digging into this more myself in the coming week, but I wanted to jot a few observations:

How Big Is Twitter Search? As Greg notes, the amount of searches happening at Twitter is a big unknown and further complicated by a few things. Are you counting searches that happen only at search.twitter.com? Or across all of twitter.com? Do you include the API calls? And do you count the huge calls made by Google and Bing for searches?

What Does Twitter Really Control? Related to that, will Twitter be able to push Promoted Tweets out beyond the ecosystem that it controls? IE, can it put Promoted Tweets at the top of search results at Seesmic — especially with Seesmic seems to have a deal with rival third-party ad system TweetUp to let that company reshape Twitter’s results? And can Twitter reshape the results at its biggest distribution partner, Google (probably not).

Bye, Bye, The Pure Twitter Stream Pie: Today’s remarkable in that the “pure” Twitter stream is becoming a thing of the past. Sure, some individual have accepted payment for tweets. But now things are becoming more industrialized. We have both Twitter itself and a third-party company with some serious history in sponsored search results ready to inject paid content directly into personal Twitter streams. How will consumers react? Will they have options to opt-out? How will disclosure happen, especially across the varied Twitter ecosystem? What about tweets that get displayed in web pages? What happens when sponsored tweets lack nofollow around links, something that already happens in some distribution situations. Will Google complain this is paid link selling? Personally, by the way, I’m not overly concerned about the commercialization. I think it was inevitable and as with search, revenue can help to ensure you get a good quality service.

This Tweet Brought To You By The Letters F, T & C: What’s the US Federal Trade Commission going to think about all this? Last decade, the FTC was instrumental in defining how sponsored search results should be disclosed. Twitter, it its blog post, calls Promoted Tweets “organic” tweets and not ad:

Since all Promoted Tweets are organic Tweets, there is not a single “ad” in our Promoted Tweets platform that isn’t already an organic part of Twitter.

Hmm, I beg to differ. Sure, these might be “normal” or “organic” tweets in origin, but the fact of elevating them above other tweets is what turns them into ads. That’s the core of how the FTC defined search ads back in 2002:

Because search engines historically displayed search results based on relevancy to the search query, as determined by algorithms or other objective criteria, the staff believes that consumers may reasonably expect that the search results displayed by individual search engines are ranked in accordance with this standard industry practice – that is, based on a set of impartial factors. Thus, a departure from the standard practice, such as a search engine’s insertion of paid-for placements in the search list, may need to be disclosed clearly and conspicuously to avoid the potential for deception.

Thus, any Web sites or URLs that have paid to be ranked higher than they would be ranked by relevancy, or other objective criteria, should be clearly labeled as such using terms conveying that the ranking is paid for. In the staff’s view, to avoid deception such labels need to convey that the sites listed are placed higher, or otherwise presented more prominently, because they have paid for their ranking or position, rather than solely based on some objective criteria relating to the probable relevance of their content to any particular search request.

Now, Twitter is clearly labeling these, though it uses the language of “Promoted” rather than the far more common term of “Sponsored” that the major search players settled into after the FTC’s guidelines were issued.

Personally, I’d prefer the term sponsored. It’s what we’re used to in the industry. I wasn’t happy when Google started blurring things by using Promoted – then Sponsored — then Promoted again for ads at YouTube. I hate to see Twitter further add to confusion.

For more on this, see my past post, Google Experiments With Paid Inclusion & Does “Promoted” Meet FTC Guidelines?

In addition, John Battelle had some interesting thoughts on the commercialization of the Twitter stream. Over at OneRiot, they have an interesting post on how they’re six-month old system to place ads against real-time search results has been working. That system has sounded pretty impressive in concept, though the key challenge has remained picking up the volume.

Also see related news at Techmeme.

VigLink is in the news today primarily because the company has announced that it has closed a round of seed investment, for $800,000 in total. From the press release, funding sources include:

Google Ventures, and individual investors including LinkedIn Founder Reid Hoffman, longtime Google executive and current LinkedIn VP of Product Deep Nishar, and noted technology entrepreneurs Niel Robertson, Hadi Partovi, Ali Partovi, Carlos Cashman, and Micah Adler. The financing, which was co-led by First Round Capital and Google Ventures, will be used to accelerate VigLink’s market expansion.

At Search Engine Land, we don’t tend to focus on affiliate programs. But I felt VigLink was worth a closer look because of the Google involvement and because of questions the company’s program raises in terms of Google’s guidelines about paid links.

Google’s involvement is straight-forward. Google has a venture capital arm, Google Ventures. The company likes VigLink’s prospects enough that it has decided to back it with funding (how much isn’t disclosed).

On the paid links front, my understanding is that VigLink passes Google’s paid links requirements. I’ll get into that more, but first, how it works.

Install Code & Get Affiliated

VigLink provides code to those who are in its program. Similar to code used for something like Google Analytics, you insert this so that it’s present on every page.

The code reports to VigLink all the outbound links on your site — or specifically, all the outbound links that people are actually clicking on. If notes these links and checks to see if the companies the links lead to offer affiliate programs. If they do, then VigLink turns those links into affiliate links that earn you money.

VigLink’s demo page gives a real example of how this works. On that page, there’s a link to a book at Amazon called “Four Steps to the Epiphany.” The link in the page looks like this, if you hover over it:

http://www.amazon.com/Four-Steps-Epiphany-Steven-Blank/dp/0976470705/

Click on that link from the demo page, and you’ll see how it looks like this in your browser’s address window:

http://www.amazon.com/Four-Steps-Epiphany-Steven-Blank/dp/0976470705/?tag=vig-20&sid=151gu6ptz593e

That part I’ve put in bold? That tells Amazon that this person came from VigLink. If they buy something, then VigLink earns some money off the sale, since VigLink is an Amazon affiliate through the Amazon Associates program.

VigLink has enrolled itself in thousands of affiliate programs similar to those that Amazon offers. Anyone using VigLink code on their web site is effectively becoming part of VigLink. VigLink would look at your site, and if it sees a link to Amazon that doesn’t have an affiliate code on it, then it would add a tag to make it part of VigLink. The same is true for other online retailers such as eBay or TigerDirect.

Visitors to your site see none of this. If they hover over a link, they won’t see the affiliate code. Nor does VigLink operate like some of those programs like Vibrant that automatically try to make any word on your site. Only links that you’ve already created are turned into affiliate links, assuming there’s a relevant program.

Advantages

The plus side is the time savings. On my personal blog Daggle, for example, I link to things all the time. If I’m really thinking about it, I might add my Amazon affiliate code to a link to relevant information in something I’m writing. If I’m busy, I probably won’t bother. Other companies I link to might have affiliate programs, but it’s not worth my time or effort to sign up for those.

VigLink aims to correct such laziness. For example, I’m in the closed VigLink beta right now. I can see that in the week or so since I put the VigLink code up that I’ve had 53 clicks from my site over to Amazon. Of these, 47 clicks were on links where I had added affiliate codes. The other 6 clicks? I never added the code, but VigLink did it for me, using their code.

Similarly, I had nine clicks from my site to TigerDirect. I’m not a TigerDirect affiliate member. I have no affiliate code on those links. But by adding VigLink, it added its own code to those TigerDirect links, making them earn for me.

Now, I’ll earn more if I do affiliate links direct. VigLink takes a cut of any affiliate sales (how much varies and isn’t specifically disclosed, to my understanding). If you’re really worried about earning as much as possible, you’d want to do your own direct affiliate links. And if you do, VigLink recognizes these and doesn’t replace them, the company says. It only act on links where it sees a failed monetization opportunity.

Disadvantages

On the con side, the biggest is probably disclosure to your readers. When I affiliate a link, I usually make a note of this to my readers — and that’s something recent FTC guidelines now say is a requirement. VigLink suddenly makes any link on your site into one that requires disclosure. But since you don’t know which ones — and right now, since you can’t selectively choose which links might be affiliated — to cover yourself, you probably should have a blanket disclosure statement as part of your articles.

Another con is that the program doesn’t automatically find words or products that probably should be affiliated. If you write about a product but don’t link to it at some site, you’re not going to earn. In contrast, a program like Google AdSense tries to automatically look at your content and put ads near your content. And a program like Vibrant will actually turn words in your stories into links. Of course, many people find that type of word-to-link transformation annoying, myself included.

Paid Links & SEO Issues

From a search engine optimization perspective, VigLink seems a minefield. Google wants any paid link on a web site to be blocked from its crawler or flagged with a nofollow attribute, lest the site owner face a paid links penalty. VigLink potentially makes any existing link into a paid one, and site owners don’t know which, nor can they control the implementation.

Since VigLink uses JavaScript, there’s the potential that if Google does NOT read the JavaScript, it won’t see the paid links. That would mean the site is effectively getting a cloaked paid link. The search engine sees a non-paid one, but humans see a paid one. Arguably, this could be a guideline violation. Then again, since the site owner may not have deliberately set out to place a link in return for payment, Google might not view it that way.

If Google DOES read the JavaScript, as it does in some cases, then some of the links are definitely transformed into paid links — and currently, without any nofollow attribute attached to them, VigLink tells me. Some of them probably do not have redirect blocking that would make up for the lack of nofollow, so again, there’s a potential guidelines violation.

Google Gives All Clear

VigLink said it has talked closely with Google and been assured that its program will not cause site owners to violate paid link guidelines. I’m also waiting for a longer statement from Google. But Google also reassures me that it has seen the VigLink program, and it doesn’t pose any problems.

Other Issues

Another search marketing concern is the lack of control you have over where your site’s reputation can flow. Some sites want to ensure external sites receive get PageRank or reputation credit that’s typicaly passed through links. A program like VigLink, if done to Google’s paid link requirements, means that this credit is blocked. This isn’t the case within domains, by the way. VigLink says it never changes a link within the same domain or subdomain of a hosted site.

Long-term, if VigLink is successful and attracts imitators, the links that Google depends upon for much of its relevancy would be lost, since it wants to ignore any that are paid. Of course, affiliate links aren’t a new issue on this front. In the past, Google has suggested that there’s always enough of a sample of non-paid links for it to do good link analysis.

Ideally, I’d like to see site owners have much more precise control over what links get affiliated and which don’t. I’d also like to see VigLink insert nofollow attributes to links that are processed through JavaScript. Similarly, I think it should consider placing nofollow around any links on a page that it may make paid when the JavaScript processes.

CEO Oliver Roup told me last week these are all things that might happen in the future. But on the nofollow front, right now it’s not so pressing, since Google’s given the program a clean bill of health as things are now. As for granular control, VigLink says that if that comes as needed during the beta, it’ll happen afterward.

Of course, much of this is moot right now. VigLink is in a closed beta. You can’t sign up for it. You can, however, generate what’s called a “pre-sales” report. This is designed to show you links on your site that could be monetized by VigLink when it opens generally in the future.

Postscript: James Morell alerted me to Skimlinks, which seems to operate in the same way as Viglinks. I haven’t looked at Skimlinks in any great detail. In a quick visit, one key difference between it and Viglink is that Viglink says members get paid each month, regardless of how much they earn. Skimlinks says there’s a roughly $160 minimum earning needed in a month (50 UK pounds).

Here are some verbatim excerpts from the SEC filing:

Negotiation and Execution of the Definitive Agreements

Pursuant to the terms of the Letter Agreement, the parties will negotiate and execute the Definitive Agreements as soon as practicable but in any event by October 27, 2009 (the “Negotiation Period”). If the Definitive Agreements are not executed during the Negotiation Period, the parties will submit any disputes regarding the final terms of the Definitive Agreements to an arbitration panel.

Conditions to Commencement and Termination Prior to Commencement

Prior to the Commencement Date, the Letter Agreement and Definitive Agreements may be terminated only by (a) mutual consent, (b) if a breach renders a condition incapable of being satisfied by the Termination Date (as defined below), or (c) if the conditions to commencement have not been satisfied by July 29, 2010 (the “Termination Date”); provided that Yahoo!, in its sole discretion, has the right to extend the Termination Date by six (6) months if the required antitrust approvals have not yet been obtained.

Search and Advertising Services and Sales Agreement

For a period of ten (10) years beginning on the Commencement Date (the “Term”), Microsoft will be Yahoo!’s exclusive technology provider for algorithmic and paid search services and Microsoft will provide contextual advertising to Yahoo! on a non-exclusive basis. Yahoo! will be the exclusive worldwide relationship sales force for Yahoo!’s and Microsoft’s premium search advertisers.

The services provided by Microsoft under the Search Agreement will be provided on all web sites, applications and other online digital properties owned or operated by or on behalf of (a) Yahoo!, Yahoo! subsidiaries and Yahoo! joint venture relationships, as well as on software applications developed or distributed by Yahoo! or Yahoo! subsidiaries that provide access to or enable algorithmic search services or paid search services (“Yahoo! Properties”) and (b) Yahoo! Syndication Partners (as defined below), as well as software applications developed or distributed by Yahoo!’s Syndication Partners that provide access to or enable algorithmic search services or paid search services from Yahoo! (“Syndication Properties”). “Syndication Partner” means a third party with whom Yahoo! has contracted to provide algorithmic search services or paid search services.

Subject to certain specified restrictions, Yahoo! will have full flexibility with respect to the user experience, content and look and feel on all of its web pages, and will also be entitled to use the paid search services and algorithmic search services for non-internet search queries with minimal restriction . . .

Microsoft’s mapping services and mobile search services. Yahoo! may implement each of the mapping services and the mobile search services on a non-exclusive or an exclusive basis. Yahoo! also has the option to work with Microsoft to implement the services on other platforms. If Yahoo! elects to receive services for other platforms, it must receive such services on an exclusive basis.

Revenue Share Payments and Other Payments

During the first five years of the Term, Yahoo! will be entitled to receive 88% of the net revenues generated from Microsoft’s services on Yahoo! Properties (the “Revenue Share Rate”). Yahoo! will also be entitled to receive its share (at the Revenue Share Rate) of the net revenues generated on Syndication Properties after the Syndication Partner’s share of net revenues is deducted. For new Syndication Properties during the Term, and for all Syndication Properties after the first five years of the Term, Yahoo! will receive its share (at the Revenue Share Rate) of the net revenues generated from Microsoft’s services on Syndication Properties after the Syndication Partner’s share of net revenues and certain Microsoft costs are deducted.

On the fifth anniversary of the Commencement Date, Microsoft will have the option to terminate Yahoo!’s sales exclusivity for premium search advertisers. If Microsoft exercises its option, the Revenue Share Rate will increase to 93% for the remainder of the Term, unless Yahoo! exercises its option to retain its sales exclusivity, in which case the Revenue Share Rate would be reduced to 83% for the remainder of the Term. If Microsoft does not exercise such option, the Revenue Share Rate will be 90% for the remainder of the Term.

Microsoft will also pay Yahoo! a payment of $50 million annually during the first three (3) years of the Search Agreement. Yahoo! may use these payments to partially cover transition and implementation costs not otherwise covered under the Search Agreement.

Termination Provisions

Yahoo! may terminate the Search Agreement if the trailing 12-month average of the RPS in the United States (the “U.S. RPS”) of Yahoo! and Microsoft’s combined queries falls below a specified percentage of Google Inc.’s (“Google”) estimated RPS measured on a comparable basis or if the combined Yahoo! and Microsoft query market share in the United States falls below a specified percentage; (d) on the fifth anniversary of the Search Agreement, and any time thereafter, Yahoo! has the right to terminate the Search Agreement if the trailing 12-month average of Yahoo!’s U.S. RPS is less than a specified percentage of Google’s estimated RPS; or (e) subject to exceptions, either party may terminate if a law, regulation or order would have a significant, adverse impact on a primary aspect of such party’s intended benefit of the Search Agreement.

Transition and Implementation Plan

Microsoft will hire not less than 400 Yahoo! employees (the “Transferred Employees”) and will offer the Transferred Employees market competitive compensation packages. In addition, Yahoo! and Microsoft will mutually agree on a retention plan to be paid for by Microsoft to assist in retaining the Transferred Employees and an additional 150 Yahoo! employees to be mutually agreed upon between Microsoft and Yahoo! to assist with providing the transition services.

The deal seems broader than the “web, image and video” search scope we heard before. Yahoo can terminate the deal if certain targets aren’t meet surrounding revenue per search, benchmarked to Google. At the five year mark the parties can change who runs “premium sales.” The rev share percentages change accordingly.

On a related note, there’s an opinion piece in the Wall Street Journal that argues these terms give Redmond a bigger win than Yahoo in Microsoft CEO Steve Ballmer’s “win-win” scenario but that, in the larger scheme of things, Google may have already won the game:

The deal is a clear win for Microsoft and a qualified win for Yahoo. The big question is whether it makes any difference in the only contest that really matters, which is the one with Google. The risk for both Microsoft and Yahoo is that the contest is already over. Second place won’t really matter, especially as the competition shifts to Microsoft’s home turf: operating systems.

Related coverage:

• It’s Finally Official, Microsoft & Yahoo Make A Deal, Yahoo Gives Up On Search
• Live Blogging The MSFT – YHOO Search Press Conference
• Microsoft-Yahoo Deals 2008 & 2009, Side-By-Side
• A Search Eulogy For Yahoo
• Micro-Hoo Details: Q&A With Mehdi & Schneider
• The Microsoft-Yahoo Search Deal, In Simple Terms

Here’s the issue in a nutshell. According to a March 2008 survey by Truste (written up in MediaPost), “nearly three out of four people, or 71%, said they realize that companies track their Web browsing activity for purposes of sending them targeted ads. The majority–57%–said they are not comfortable with the practice, even when their browsing history can’t be linked to their names. At the same time, 72% of Web users also told researchers they find irrelevant ads ‘intrusive and annoying.’”

The US Federal Trade Commission (FTC) held hearings on personal data collection and last year proposed voluntary rules for ad targeting on that basis. Just this week the IAB adopted formal guidelines around consumer privacy and when users should be notified and given the opportunity to opt-out of targeting. The IAB’s rules are apparently more “flexible” for publishers and advertisers than the FTC’s and don’t require as many consumer disclosures and notifications.

Every BT firm and practitioner out there will say that it works effectively without personally identifiable information. However, there are many in Washington and Europe, as well as privacy NGOs, who simply don’t buy those claims.

Earlier this year ISPs in the US and UK began collecting data on user behavior for resale to ad networks and others. In the US a firm called NebuAd provides the enabling equipment that captures the data and permits it to be shared with third party advertisers and networks. In the UK NebuAd competes with a company called Phorm. And there are others seeking to bring more targeting capabilities from deeper data mining of user profiles and activities. Yahoo is on the forefront of this in the US.

Yesterday, however, cable ISP Charter Communications backed away from a plan to work with NebuAd, citing privacy issues. According to the Washington Post:

The company had been planning to harvest the stream of data from each Internet customer for clues to their interests and then make money from advertisers who would use the information to target online pitches.

The data-collection effort would have protected personal information, Charter officials said in describing the plan, but critics likened the practice to wiretapping.

Charter said the plan had indefinitely been put on hold. This week, in a parallel but contrary development, the US House of Representatives passed domestic spying legislation that enables federal officials to obtain copies of all communications (phone or Internet) in the US without first obtaining a search warrant or court order.

Privacy advocates see little difference between what firms like NebuAd are doing and the efforts of the US government to spy on citizen communications. It’s therefore somewhat ironic that the US FTC is asking private companies to offer broad disclosures about data collection when the government is adopting an entirely different standard for its own activities. (The justification for the double standard is criminal law enforcement.)

On the advertising side, the question of who “owns” the data and what the scope of protection for that data should be will continue to be the subject of debate for some time. In Europe, it has been proposed that an IP address should be treated as personal information. That position has all sorts of privacy implications that fly in the face of BT and other targeting initiatives. Indeed, most advertisers and ad networks argue against such a policy.

Investor Esther Dyson, in a related piece in the Wall Street Journal this past February, wrote that a “coming ad revolution” is tied to individual management of personal data (on social networks). She also supports NebuAd and related firms in their effort to provide more relevant ads to users. Dyson’s bullishness has a one-sided, naive quality about it, however.

Facebook’s tracking initiative Beacon failed because it was manipulative and somewhat deceitful. Many users didn’t know their behavior was being rebroadcast to their networks. When they found that out, they were upset in many cases. Facebook was forced by bad PR and related negative coverage to drop the program, just as Charter Communications made a similar calculation before the PR problems could begin.

The issue of data mining and ad targeting is extremely complicated. The genie is out of the bottle, however, given that these capabilities exist. And users do prefer more relevant ads. In this environment the challenge is to find the right balance between data collection, discloses, and consumer choice around being tracked.

And that’s not an easy thing for anyone.

At the recent SMX Advanced event in Seattle, I sat in on the Amazing New PPC Tactics panel. It was a good panel overall, but the one thing I wanted to highlight was the presentation by David Szetela of Clix Marketing, as David explained how to do content match marketing effectively.

Understanding the problem

The problem stems from the way that the search engines decide on which sites to run your content match ad. Basically, the only clues that the search engines have is the list of keywords you provided. What would be wrong with that, you might ask? Let’s illustrate with an example.

Let’s say you are selling a variety of widgets, and that these widgets can be used in your kitchen, on boats, in your workshop at home, and also has applications in industrial machine shops. You might create a list of keywords that looks something like this:

kitchen widget boat widget workshop widget machine shop widget

These are the keywords that the search engine will then use to try and match your content match campaign up with web sites for placing your ads. With the above list you might end up on sites about kitchens, boats, workshops, and machine shops. The reason this happens is that the search engine takes the keyword set and looks for sites that have similar words on their pages.

This is significantly different than the way the keywords operate in search. In search, the search engine is trying to match you up with a user’s search query—with an average of only about 2.3 words per query. Not a lot of data to match up with, and usually pretty focused on nature. As a result, with search campaigns, there is a tendency to create as large a keyword list as possible, to cover a lot of potential ground.

This may work in search, but in the content match world you can rapidly get into trouble. In our example above, you may not want to be on sites about kitchen design, or boat vacations, etc. Worse still, you may not want to be on a site about travel that happens to have a few pages about cruises you can take (e.g., if your widget helps people repair boat engines, this would not be a match).

The solution

It turns out that I have always had that first step right. Turn off content match in your traditional campaign. It’s just going to be very tough to do well with it.

Then, create a new campaign, and turn off “Google Search” and “Search Network” in your campaign settings, and create a campaign which is only for content match. This is the second step.

For the third step, create a custom set of keywords just for your content match campaign. These keywords will be significantly different from your search keyword set. Here’s some of what David recommended we do to pick out the keyword set:

1. Pick out a few sites that are strong examples of where you would like your ad to run

2. Do an analysis of the types of phrases that appear on these sites (which is what the search engine will do)
3. Consider going further and entering some of those phrases in the search engines to discover a few other sites you would be interested in being on, and expand the analysis
4. Limit the list of keywords to something between 15 and 30 keywords

With this approach you will be significantly closer to targeting your content match campaign. Better still, since many people simply turn this feature off, and most others use their search keywords, you will have a better optimized campaign and a significant competitive advantage. David offers many other Content Match Optimization tips on his site.

Eric Enge is the president of Stone Temple Consulting, an SEO consultancy outside of Boston. Eric is also co-founder of Moving Traffic Inc., the publisher of Custom Search Guide.

Yahoo will be powering the ads within PDF documents that are contextually relevant to the content within the PDF. If you produce PDF documents and want to monetize them with relevant ads, this seems to be the product for you.

To get ads on your PDF documents, you will first need to be approved for the beta. To sign up for the beta, complete this form. Then you apparently upload your PDF documents to the system, then Adobe and Yahoo will place the ads within the documents automatically.

Here is a sample image of Yahoo ads in Adobe PDFs:

The PDF FAQs has more information on how this all works. Here is a copy of the PDF document as text:

What is Ads for Adobe PDF powered by Yahoo!?
Ads for Adobe PDF is a new opt-in service that enables commercial publishers to monetize their PDF content through advertising. The advertising service, powered by Yahoo!, automatically matches and displays dynamic, contextual ads in a panel of Adobe Reader and Adobe Acrobat when a user views the PDF. Commercial publishers that sign up to participate in this program are eligible to be paid for valid clicks on ads. In this unique advertising relationship, Yahoo! helps connect advertisers from its extensive network of online advertisers with commercial publishers that are distributing content through Adobe PDFs. Yahoo! will incorporate this new advertising inventory into its portfolio of online advertising opportunities.

How does it work? When a recipient opens a PDF document from a commercial publisher participating in this program, ads will be dynamically matched and displayed in a panel adjacent to the content. The publishers are eligible to be paid for valid clicks on the ads. Adobe has partnered with Yahoo! to provide ads across a broad range of categories to match PDF content. Since the ads come from Yahoo!, publishers do not have to establish and maintain direct relationships with advertisers.

To use the service, publishers follow these easy steps: 1. Apply for an account. 2. Upload and register the PDF content. 3. Distribute the registered PDF content – just as they do today.

PDF registration associates the publisher with the content in order to receive payment and automatically analyzes the content to enable ad matching.

How will ads be matched to PDF content and will they be relevant? The PDF content is thoroughly analyzed, using an algorithm that takes into consideration factors such as: * Keywords – identifies keywords based on how frequently the word appears, where it appears in the content, and how it appears in the content, among other things * Context – associates categories or concepts with specific words; for instance, content about soccer could be associated with sports

This analysis enables contextual ads to be matched to the PDF content.

How is this different from ads placed in PDF content today? For commercial publishers, the current approach for securing advertisers or sponsorship for PDF content is very time-consuming. Ads for Adobe PDF takes advantage of Yahoo!’s existing relationships with an extensive network of advertisers, and provides publishers the opportunity to monetize PDF content with little effort and without the need for an advertising sales force.

For readers, Ads for Adobe PDF provides dynamic ad content that is up-to-date and related to their interests. Given this new opportunity to monetize PDF content, some commercial publishers may choose to provide readers with access to more free content, which previously was available only to paid subscribers.

For advertisers, Ads for Adobe PDF provides a powerful new channel to communicate their messages to a highly engaged and targeted audience. In addition, Ads for Adobe PDF provides advertisers a means to track the performance of ads placed in PDF content, just as they can today with ads placed in web sites. That’s an important distinction from current ads in PDF content, which are static, and offer no means to measure the effectiveness of the ad placement.

When will Ads for Adobe PDF be out of beta? Adobe and Yahoo! anticipate several beta cycles prior to release for General Availability. We will provide updates on the release schedule in the coming months.

• Facebook Pages allow brands to have profile pages on Facebook in the same way that people have them. (see Coca-Cola’s screenshot from ZDNet) with marketing messages, games, video, etc. When users interact with those pages, add elements to their profiles or brands as "friends," that information is spread virally through the Facebook newsfeeds, just as it is today with the Facebook apps. A related program, Insights, will offer very specific analytics and data to marketers, in the aggregate.
   
• Facebook Social Ads allow brands to use the full range of profile information to target people by age, gender, relationship status, location, and so on. MySpace launched a conceptually similar ad program "hypertargeting ads" Monday.
   
• Facebook Beacon is a widget-like system being adopted on 44 third party partner sites (e.g., Blockbuster, Hotwire, The Knot, TripAdvisor, TypePad, Yelp, among numerous others). Activities on sites using Beacon will flow back into a Facebook user’s News Feed or Mini-Feed, allowing their friends to know some of what they did at that site. This effectively represents a way for individuals to "endorse" brands, products, or businesses on partner sites and broadcast that information back to their networks on Facebook.

On the one hand, this set of announcements represents a creative, clever, and logical use of the Facebook demographic and profile information and the Facebook platform. Beacon is sort of like Facebook apps in reverse. Both represent ways for marketers to attempt to reach consumers through Facebook’s much touted "social graph" (awful term).

Brand advertisers are hungry for effective ways into social networks and to reach their generally young user bases. The potential problem here for Facebook is that it may be moving a little quickly and may be pimping its users a bit too aggressively. Those who feel they’re being watched, used, or manipulated somehow by these ads may resent it. Privacy is a potential concern as well, though Facebook has sought to address that by using anonymous, aggregated data for targeting and reporting.

Facebook Platform was a stroke of genius, but I find that much of what now goes on at Facebook (via the apps and their adoption) qualifies as a kind of spam, even though it is pushed via my network. Clearly not all people share this view. But in my opinion, while Facebook Ads may make the site engaging and safe for brand marketers, it may start to become much less so for regular people.

For more, see the press release from Facebook and the Facebook blog post. TechCrunch live blogged the event, and there’s plenty of coverage on Techmeme.

Marketers with existing AdWords accounts may create ads within BuzzLogic for the particular sites where the “conversation” about a product, service, or brand is happening. This capability doesn’t exist within AdWords/AdSense itself. BuzzLogic has said that other ad networks will follow in the near future.

What this means is that marketers can locate, create, and then place ads, say, for digital cameras, on blogs that have the greatest influence (as determined by the BuzzLogic algorithm) on the online discussion of digital cameras. This has many obvious and some less obvious but important implications for brands as well as direct marketers. As social media and blogs gain in influence, marketers are seeking ways to take advantage of or combat their influence. What’s interesting about BuzzLogic’s tools is that they make that influence much more transparent to marketers. And the new ad targeting options make it actionable.

Other blog search engines and tracking tools, such as Technorati, allow audiences and marketers to see blogs that may have overall influence or are discussing particular topics. But they don’t rank blogs for influence on particular topics or keywords, let alone allow marketers to buy ads on those specific sites.

BuzzLogic said its beta advertisers saw conversions that were as much as 2x better than other approaches.

Superpages is charging advertisers a one-time $990 production fee and a small monthly recurring fee to have a video icon associated with their listings. For each consumer stream/click, advertisers would also be charged a fee — pay per stream — based on competitive bidding on a category basis. Category level bidding is the same way that Superpages’ clicks and calls products work.

Advertisers can also pay a one-time fee for the rights to use the videos on their sites or distribute them elsewhere — on or offline. Superpages is also considering various syndication options for the video, similar to the way it syndicates calls and clicks. Syndication from yellow pages publishers (and other local sites like Citysearch) will likely be the way that local and SMB video gets into search or local results at Google, Yahoo, and Microsoft.

Here are a number of examples of Superpages video in different business categories.

Video is popular with both consumers and advertisers, as lots of empirical work already shows. A recent online survey from Opus Research and AllBusiness.com of 653 U.S. SMBs found that 33% of respondents said they were evaluating developing video ads in the next 12 months.

There’s some additional discussion of the particulars of the Superpages program on my personal blog, Screenwerk.