As Google rolls out “enhanced” campaign setup options and parameters, various accounts of the changes are referring to some backlash from concerned advertisers. We’ve even been notified of another Change.org protest of the (yet-to-be-experienced) new features.

It’s to be expected, given how competitive the environment is and how hard many of us work to turn a profit for our clients’ campaigns, that suspicion abounds that this is just all about Google making more money from mobile.

But if you listen to Nick Fox’s account of it, it syncs up with the idea that the current approach is simply too cumbersome. Big accounts are indeed “pivoting around” way too many dimensions. In another year’s time, many of these accounts would go from having “way, way too many campaigns” to “way, way, way too many campaigns.”

Certainly, there is an economic question in this. But rather than it being about “Google making more money from mobile,” Google is looking longer term and simply trying to stem a potential loss of growth in revenue overall. Indeed, excluding emerging international markets, if the shift to mobile usage patterns accelerates, and the number of clicks on desktop computers declines sharply, Google could make less money year over year. Can you imagine that? One day — despite whatever it might do to help advertisers better control mobile ads — that could in fact happen. As newspapers and network TV are finding, advertising isn’t forever.

Google is actually making a lot of money now from inefficient mobile deployment. Contrary reports, you don’t need to set up separate campaigns to bid separately on mobile ads, or to activate ads in that channel. They’re on by default. So this change is likely to be revenue-neutral, or even revenue-unpredictable. By reducing advertiser frustration with the channel (and unfortunately, “sexing up” the change with the usual examples of extreme bidding precision based on proximity, time of day, and whether the prospect has showered that day), Google seems to be looking long term.

As for the collapse of desktops and tablets into a single channel, we largely collapsed these back together this year after experimenting with splitting them out. We feel that they perform similarly enough as to not warrant additional “pivoting” around yet another dimension. My hunch is that this shift — essentially, to call tablets desktops — is something Google hopes will slide by analysts in the maelstrom of data and forecasts. By defining tablets as “computers,” they’ll show a less horrific decline in clicks from “computers” from quarter to quarter, in their financial reporting. That will spook Wall Street less. And for some reason, Google has reached a stage in its history where it seems to be very concerned about any optics that might cause a downward shock to its stock price. A far cry from the company’s insouciant attitude of days gone by.

Today, I’ve been going through all of our client AdWords data, looking at the aggregate CPM rates Google is commanding for both Search and Display advertising. CPM (cost per thousand impressions) is a lesser-used metric in our field today, but it can provide additional perspective as a benchmark.

That exercise reminded me that many of the dominant forms of “display” advertising in client accounts today are in the various types of remarketing. That makes aggregate figures misleading, because remarketing pulls the average CPM pricing up.

The more these higher-value forms of display advertising crowd out the lower-value stuff, the happier Google is, financially. Could it be that this will also help the publisher ecosystem? One hopes.

Just a back of the envelope calculation, but here is roughly what I’m seeing:

• The strongest forms of remarketing are priced at 10X (or more) what display advertising is typically selling for in the case of advertisers not being particularly performance-focused on the latter
• For less valued forms of remarketing, the ratio is still 5X
• Even when advertisers are managing tightly to some performance metric (not as demanding as search, but tight) in display advertising, remarketing typically goes for 3X the price of the other display advertising, and at least 2X

Here’s where advertisers need to be careful. Remarketing is harvesting existing demand, leveraging past investments, and speaking to certain kinds of audiences. It’s “low hanging fruit.” That’s why it’s worth spending the money. But it shouldn’t be confused with a serious effort to build out display advertising. Many of us believe that advertisers under-invest in display advertising compared with peers who do it well. Remarketing isn’t really the same thing. It’s good, but it also shouldn’t be an excuse to get complacent. If you can find great prices on the regular display inventory, then there may indeed be a benefit.

As for Google, and publishers, this must be good for profit margins. There is too much near-worthless display inventory out there. Getting some of those clicks to shift over to audience-based targeting is a better way to monetize user attention… financially speaking only, of course.

BlackBerry will be one of the companies lining up for the privilege of paying several million dollars for a brief advertising slot at the Super Bowl.

Chances are that ad, despite its fleeting nature, might rival the company’s digital ad spend for an entire year in a country like, say, Canada.

Are there really any drawbacks to this? Unlike most TV advertising, these ads become something of a cultural touchpoint, get huge exposure, get additional views online during and after the event, and garner additional references by the tens of thousands as “earned media.” The $4 million or so won’t exactly put a huge dent in the company’s cash reserves (still in the billions).

In light of that, it could be argued that it’s good bang for the buck on paper, and in this case, particularly timely. It’s almost tempting to come around to the “pro-Super-Bowl-commercial” line of thinking.

Ultimately, though, it might pose a risk in two ways.

1. If there isn’t anything truly new or remarkable in the commercial, then you’ve reached out to the 99% of the phone buying public who simply aren’t rabid BlackBerry fans, giving them ever more reason to bray about the new devices’ perceived shortcomings.
2. It can serve as a distraction from the much more important challenge: the much more broad-ranging, “on the ground” ebb and flow of outreach across many channels, in every conceivable forum for public opinion formation & info gathering.

The second is the most important. Despite the hype, people don’t really get their information from Super Bowl commercials. The “real action” is all over the place. Especially online.

People stopped getting their information from TV spots a very long time ago. And starting 15 years ago, folks in rarefied fields began using news aggregator tools (or Bloomberg terminals) to find stuff out. Now everyone looks everything up on a constant basis, joins the fray, exercises intent by opting into a notification list or buying immediately, etc.

Is there anyone in the market for a new smartphone who is not now aware that Blackberry’s new devices and new OS are out? Is there any particular value to building the “brand” with a steady onslaught of TV and billboard ads, for example… as opposed to engaging in all other forms of more direct marketing and customer care?

Imagine if you could do for TV what you could do for some online channels: “negative out” certain demographics or intents, to reach a subset of more likely candidates. Or go after certain IP addresses that betray a likely business user. Let’s face it — when it comes to large-audience, network broadcasts,  TV’s concept of targeting is laughable. There are still folks who think you’re “targeting a business audience” by advertising insurance, enterprise software, and Cadillacs on the golf broadcasts (clearly the province only of top execs).

Well, you can’t do the “negative out thing” or the “B2B targeting thing” on the Super Bowl. So you’re paying $4 million to show a 30-second propaganda film to everyone instead of much less to reach out with a richer piece of information for some key decision-makers in the enterprise. But you knew that.

What of the mass commercial market? Not much better.

A commercial has very little connection to what we see today on BlackBerry’s blog, for example, where loyalists rush to the defense of the new Q10 devices (the ones with the physical keyboards coming out in a few weeks). What a blessing for BlackBerry that they still have a not-inconsequential tribe of people eager to jump on a comment thread with comments like (real comment): “The BlackBerry physical keyboard and call buttons are what silverware is to eating food. Yes, I can eat with my hands, but it’s much more civilized to eat with silverware.” And of course there are a lot of laments about the shape and size of just about everything. In short, a lot of real-world back and forth.

Ha! I’m guessing that’s not how the Super Bowl ad will be. It will — shocker — try to wow us with how good the device is. And in a world of people who know very well that information doesn’t come from commercials… that makes it pretty much irrelevant.

Let’s hope the company thrives down in the weeds, where the real action happens. Like here:

Getting the story out one user, one search result, one tweet, one carrier, one feature rebuild, one developer, etc. etc. etc. … at a time. By contrast, a Super Bowl ad — while it seems BIG — is easy to do. A little bit expensive, but really nothing to quibble about. And such an easy check to write.

As such, I can’t see it making any difference. Most viewers will still be talking about the Oreo ad the next day — or whichever one was really the best.

The pointless 30-second spot neatly does solve the problem with so many forms of non-traditional advertising… there are a lot of cheques to write. For many harder-to-execute feats. Sometimes the available “advertising” inventory in a certain online channel is much smaller than your wallet is (or maybe you just aren’t looking hard enough at how to spend it). When your budget is big and timelines compressed, the margin of safety says, just in case, write that big check and get it over with.

No harm done (unless the ad is laughable and sinks your brand in one day). But the other stuff — the harder stuff with all those little to-do’s and little checks to write — is what really matters.

What are your thoughts?

Today’s announcement of Facebook Graph Search looks promising.

From an advertiser and investor perspective, narrow though that may be, it suggests that Facebook will indeed add a compelling and intuitive offshoot of functionality that it can effectively monetize. That’s especially important as Facebook learns to monetize in a more intuitive way; many efforts to date have been awkward.

The limitation here is that for this to contribute incrementally to profit, people will have to spend more time on Facebook than they already do. User attention is finite — there are literally only so many waking hours that people can spend looking in the general direction of content that might have ads nearby. So really, how much upside does this create for Facebook, financially? Some. Not lots.

Without having a chance to play with the tool, it looks from here like it will be fun and easy to use, leveraging some state-of-the-art natural language capabilities. It will put some power and creativity in users’ hands. It will also be in tune with the multiplicity of forms of content — written, visual, audio, etc.

It amounts to Facebook refining and improving the experience for those who are already committed to life inside its walled garden.

The major threat to Facebook is not now, and has never been, that it can’t get good enough at being the leading social network.

The major threat is that people’s appetite for sharing their lives and content on Facebook — the fodder for all that goes on inside — will wane or dissipate. That may not seem imminent. But all the addition of Facebook Graph Search to FB’s overall feature set does is to rest even more weight on one — potentially shaky — foundation. User trust, willingness to be open, willingness to build large lists of “friends” and confide in them as if they were the closest of confidantes. Let’s face it: the searches would look pretty stupid if you only had 4-5 Facebook friends, and if those friends were taciturn about sharing anything.

So Facebook’s usefulness is premised on people building lists of hundreds and hundreds of friends, and most of those people sharing a high volume of their sentiments, content, and behavior. That may very well continue. None of this works very well if people choose to stop friending and sharing.

After a long dormant period, the issue of paid inclusion bubbled up to the surface of news about the business of search recently.

Danny Sullivan, an anchor of continuity in an industry that sometimes suffers from fragmentation and Attention Deficit Disorder, has patiently covered this intersection of the searcher’s user experience and search engine companies’ business models since the first debates about it over a decade ago. Sullivan, not shy about invoking the need for the FTC to investigate potential bias in search results, has generally advocated fuller disclosure of search and directories’ business models. Paid inclusion, he has typically argued, should be well disclosed — presumably because many users are under the impression that search results are unbiased and scientific.

As they should be.

Search engines and categorized directory driven companies often have a vested interest in the confidence consumers have in the unbiased nature of the results. Quite simply, “real” search is distinguished from advertising; both have a role, and both become most useful when people know which is which. Many search engine companies actually feel strongly about this.

Danny also gives equal time to the notion that “paid relationships can be good,” because they raise the bar on quality control standards — and that can help protect consumers.

That being said, it’s all too easy to oversimplify the business challenge of coming up with a business model that would support a great search engine or categorized directory company — or any useful content, for that matter. And in search, analysts, consumers, and gadflies have rarely dug into the details of how different approaches work. Rather — and surprisingly — critics lazily dismiss all paid inclusion models they don’t understand as simply biased or compromised.

Missteps and Abuses Eroded Trust

For their part, the search engines and directories made huge missteps when their “revenue teams” got out in front of the “I love the product” people. LookSmart, a categorized directory, lost credibility when it began charging for both directory listings and paid clicks within those listings. Yahoo’s directory merely charged for “premium listings” in its directory, but in the surface-level analysis we heard in those days, hardly anyone asked tough questions about how they rank-ordered results, how editorial standards worked, etc.

In hindsight, one thing we do know is that there were abuses. Anytime there is a way to “dominate” search results, there will be abuses. For a “short time” of about a year and a half, a single Payday Loan company easily dominated its Yahoo category by buying nine listings. That’s like the small town where the nine “different” dry cleaners are all owned by the same company. Not ideal, but certainly, stranger things have happened.

I believe many of those reading Danny Sullivan’s analyses of these matters over time probably had trouble believing such long posts were needed to explain such straightforward matters — long posts made even longer by the liberal use of screen shots. It’s a style some other analysts began to copy, on the mistaken assumption that the style was what made those posts helpful, rather than their substance.

The fact is, to properly and appropriately explain how some search listings and paid inclusion (or paid advertising near search results) work, it might take posts even longer than Danny’s.

The Relative Simplicity of the Yellow Pages Model

Remember the Yellow Pages? It was pure pay-to-play. Not only did you pay to be in there, but you could choose very large ad units, buy for multiple service areas and yes — even advertise in multiple categories! Certainly — though this was before the days of an “algorithm” that could mediate issues of relevancy and make editorial decisions seem to vanish, as we see today at companies like Google and Yelp — there would have to be some basic standards of relevancy at play. If there are “categories” in a directory, then you probably can’t allow a big fish to buy into every category in the book — even if they have the money.

But certainly there were incentives for ego-bidding, buying multiple service categories, and multiple service areas: if someone wanted to throw their money around, a relationship manager at the directory was there to handle their needs and to keep them happy. And those relationships really did get managed over the long term. In their declining years, no one had many good things to say about the yellow books, but one thing you never heard much of was one company complaining that another one was in there “unfairly”. Because the transaction was nearly *entirely* financial, it was all taken care of behind the scenes, with the cards ultimately being held by the directory publisher: if you’re really upset about something, you’re free to buy no listing. Hope your phone keeps ringing.

[In a more contemporary context, I recall coming into contact with a woman who was Director of Editorial at Overture, the pay-per-click search engine that later became Yahoo Search Marketing. I thought to myself at the time: she was utterly unsuited to the task of sorting out who should be allowed to advertise where across a massively scaled, keyword-driven advertising auction, as most humans would be. Google saw the handwriting on the wall there long before others did, and ramped up their development efforts on what would become the AdWords Quality Score formula, a formula that will necessarily be subject to endless refinement insofar as relevance, fairness, appropriateness and user experiences need to be balanced out with Google's business model.]

So: remember back to whether there were inter-company “squabbles” directed at the “mediator” Yellow Pages company. There was no social media to complain on, and no assumption that a nice helicopter-parent-ish “user experience company” would help you get business if you were the “most relevant”. Your competitors wouldn’t be called out as “spammers” or “cheaters” because there was no such thing.

Things are much better (and cheaper) for remarkable, strong word of mouth businesses, in this regard. Yet paradoxically, they complain more about the platforms that help word of mouth spread cheaply. Perhaps it’s because those platforms have set themselves up as paragons of virtue. Those who discover that “don’t be evil” is really just code for “be 10% less evil than the old boss,” and egged on by critics like Doc Searls and the people at Adbusters, who tend to cartoonishly lampoon and reject all advertising, forget (in their disillusionment) to build viable alternatives, or to respect that trial and error and business relationships built on choice and messes of all different shapes and sizes are epistemically preferable to naive, utopian visions dreamt up by well-meaning individuals.

Benefits of Search and Reviews

These kinds of extreme positions (about how evil Google, Yelp, etc. are) too easily forget how lame things were just fifteen short years ago, as the Yellow Pages era was declining.

“Search” isn’t sacrosanct: it’s a great way to be found. Fortunately for many businesses, Google has spent billions of dollars making it a credible enough source of information and advertising that consumers *willingly* come back in huge numbers daily.

“Reviews” are helpful but not gospel. Sorting out real from fake is a vital prerequisite for forward progress in a user-generated-content sector dominated by well-regarded publicly-traded companies like Yelp and TripAdvisor. Like Google, they invest serious resources into editorial oversight and more and more all the time into developing algorithms to improve trust and relevance. That’s a good long term bet, because consumers place far more value on unbiased information than on biased information.

That being said, they’re not out of the advertising business entirely. Without revenues, they wouldn’t be businesses at all. And things have improved greatly over the years.

When Yahoo was experimenting with its premium directory listings, coupled with (for example) Inktomi-or-Google-driven keyword search results, few sought to ask about the rules for buying in multiple categories, how relevance rules were enforced (they had editorial teams, probably an inefficiently high number of them cross-subsidized by Yahoo’s inflated early profits from ineffective banner ads), or how rank order should be determined when an advertiser was in a second or third category. Few were sophisticated enough to ask. And although there were a few online forums and so on, the amazing thing is there was no Facebook and Twitter (and no slavish fear of them so that every move a company makes needs to worry every day about “what everyone is tweeting”), so you did not have a lot of deep dives or food fights over how people’s listings showed up. Needless to say, things sucked much worse then, and people knew less.

Progress, Accountability and Transparency

The key difference today is that (1) things suck much less; (2) companies like Google, Yelp, TripAdvisor, and pretty much everyone else can’t help but hearing constant streams of input and feedback, much of it negative; (3) people think they know much more about information retrieval.

The problem is — with (3) — people think they know more about information retrieval because more people are using these sites and advertising on them. Unfortunately, few know any more about information retrieval than we knew a few years ago. Most every business is in the same mode businesses have been in their dealings with Google Search since 1999: if a competitor is doing better on Google Search, then Google Search must be broken, or corrupt. It’s just far easier to point out problems than it is to design a massive system geared to please everyone.

But search and directory companies are trying to improve as they attempt to balance revenues with ever more transparency, relevance, and trust. They will have to do so with 80% technology, and 20% editorial oversight. Consistent, published rules will be helpful, and must be applied to everyone without exception. Neither inter-advertiser fairness nor the user experience are subjective matters; both can be measured.

If search, directory, and review sites don’t follow through on those objectives, then users will leave. The services will become pure advertising, as the Yellow Pages were; services that few people would voluntarily use. These aren’t monopolies. Other services are only a click away, as Sergey Brin is fond of pointing out.

On a regular basis, business owners have useful and important feedback that helps change how search (etc.) companies deal with paid listings. Sometimes there are significant problems with a large publisher like Google abusing its power, and a groundswell is needed to force change, such as this Change.org petition to allow advertisers the old ad rotation option in AdWords. (I signed that one — I felt strongly about that issue. Naturally, there will be other petitions, as there are always going to be unresolved platform complaints. I probably won’t sign the next one, because I’ve been writing publicly to a wider audience as well as communicating with Google for years, not signing protest petitions. They don’t always listen, but I believe you aren’t heard anymore if you just complain about everything constantly, or stage “protests” on matters that should be taken care of with routine, respectful input.)

But for the most part, let me go against convention by rejecting the idea that it is fair to take to Twitter every time you don’t like how a competitor shows up in Google, how Yelp’s review filter affected *just your business*, etc., like social media is some kind of “higher court” that should be used to get a specific case looked at in more depth.

The really interesting questions are deeper ones of how these pop-up problems could be dealt with by handling them consistently over the long term — like how do categories work, how does relevance work, what do users find most useful, where are there exceptions that treat some businesses unfairly when other businesses take advantage of loopholes, etc.

Like I said, even the Yellow Pages stuff wasn’t quite as simple as it looks, in its day. “Categories” had to have meaning, and listings had to have some standards. But all of these squabbles took place behind the scenes.

Does anyone, to this day, remember how Yahoo used to rank-order the businesses paying for premium directory listings? No one gives it a thought, but you can be sure that someone at Yahoo could have (or should have, or maybe did) spent a fair bit of thought trying to sort this out, so that the user was given relevant results and the listing businesses (who all paid something) were treated fairly in relation to one another. But outside of those behind-the-scene efforts, no one appears to care very much. The listing businesses suddenly care when their competitor shows up more prominently.

All the hue and cry in the world (business owner tweets, negative press by those journalists who are squeamish about profit-making publishers who also provide great resources to the world) won’t change the missions of companies like Google, TripAdvisor, and Yelp. They do things that set them apart from the Yellow Pages, travel agencies, and the (probably bribed) concierge at the hotel.

HomeStars Seeks a Better Way

And as for HomeStars, which is what motivated me to post this in the first place: HomeStars is significantly different from the many “bought and paid for” contractor referral services that have come down the pike over the years. The core mission is to highlight real reviews from real homeowners, and the business model is supported by paid premium listings, which come with certain clear rules. Any home services business is eligible to list on the site without paying. Quality control on reviews must necessarily be achieved through a combination of editorial and algorithmic means.

To bring up one example of a thorny dilemma in how to implement multiple category membership: let’s say a business (an electrician) is listed in a single category [electricians] and doing well, and wants to upgrade to be included in a second category so customers in a related area [pools, which often require wiring] also see their listing (a privilege that comes with an extra fee). In the Yellow Pages era, the decision would be binary. Either they can pay to do it (period), and editorial says it isn’t too far out of the realm of relevance, or an editorial decision is a firm “no” (locksmiths can’t list under “cupcake stores”). In most cases, the price of the listing is going to take care of the decision. The total price tag for premium listings would climb to the point where irrelevant listings would be (mostly) self-policing.

In the online version of this, with review content, the decision isn’t quite so binary. Consumers demand superior relevance and more information about the business. Business owners may complain that a competitor “shouldn’t be in their category”. There is no black-or-white solution, but there are always ways to improve. HomeStars is considering ways of adding a separate form of feedback by consumers who found that the business wasn’t appropriate for their needs, and that they were misled by their presence in a certain category. If there is enough of that type of feedback, it could form part of the ranking algorithm, or cause the business owner itself to rethink their presence in a category which is too much of a “stretch”.

In any case, similar to Yelp, Google, et al., HomeStars can’t and won’t change its long-term mission just because one business is squabbling with another. That would defeat the whole purpose. Feedback is important to help HomeStars figure out how to do a better job of, for example, enforcing relevance in categories (not just counting reviews, but figuring out who can and should show up in multiple categories and multiple service areas… a seemingly simple matter, but not nearly as simple as it sounds… much harder than it used to be at the Yellow Pages… though as I’ve already argued, some of that complexity was buried behind the scenes in the pre-online world).

Even the mighty Google listens to constructive feedback, as the Change.org petition episode showed. Other kinds of feedback, out of necessity to the larger mission, have to get ignored in the noise of “me first” style beefs, and Doc-Searls-style out-of-hand rejections of all advertising. (The family of solutions that Searls proposes to allow customers to signal their intentions, thus obviating the need for intrusive advertising, ironically, just shift power from an interruption-based medium to a lead-generation platform; ‘winning’ business via such an ‘unbiased’ platform is no less likely, in the real world, to be free of bias, nor immune to rising costs and barriers to entry. Incidentally, HomeStars has been building an early prototype of a platform in this genre, called simply “The Project Tool.” It is currently free of charge.)

Google has shown (though Searls missed it in his book by forgetting that Google launched its CPC auction with relevancy scoring in 2002, not 2000, which was a CPM-based, fixed-price prototype) that categorized and keyword driven advertising can get better if they refine and iterate. That’s all we’re trying to do — get better. You can decide for yourself if that changes the world or not.

I recently tried to envision the future of agencies with a series of posts on the Acquisio blog, and in the process of trying to do so, realized I could not.

Bringing tidy logic to any such predictive task might create a soothing, pat, “smart”-sounding narrative, but in business, deductive logic rarely bears much resemblance to real-world outcomes.

Instead of trying to “predict,” I looked at what kinds of service organizations (ad agencies, mainly) have had uncanny staying power in the rough-and-tumble of “creative destruction” in the marketplace, and saw that some very boring, uncoordinated “holding company” models were amazingly persistent in our industry. Add to that the relatively common practice of large holding companies in industries other than advertising or digital media buying up newer types of marketing services organizations as means of diversifying their holdings. Either these unusual-seeming acquisitions work out despite seeming like bad fits, or they don’t work out (as in Microsoft acquiring agencies for a very high price, later selling them to more appropriate ad agency holding companies, who presumably got to build on their strengths by picking up discarded units at a discount). Either way, evolution is at work, and the initial mistakes either lead to surprisingly robust forms of diversification, or they lead to some units being selected “out” of the gene pool. Perhaps — to push the ‘DNA’ theme a bit further — genetic mutations are also part of the picture. Say, when an organization sits for a time within the confines of a high tech company that is good at using and/or creating certain tools, maybe when it ultimately moves on to join ranks with a more appropriate parent company, that parent company nonetheless acquires some important DNA (in this case, in the form of technological capabilities that prove useful for survival down the road).

At a very high level, then, some of the “epistemic takeaways” from the “Future of Agencies” thought exercise included:

• People claim to hate “siloes” in business, yet they persist. So mightn’t they serve some useful function?
• When non-monolithic holding company organizations consider their planning, they don’t sneeze at the unsexy parts as long as they work financially. They optimize financial performance in specific units rather than trying to force a unified logic across all units. They “swap out” the worst-performing parts, after careful consideration. They “swap in” technologies and acquired companies that seem like they fit well with the long-term plan, but they don’t fold all the assets into the whole. They just let them operate, and try to improve them where they can.
• The models that work are working because of trial-and-error or experience, not “planning.” As such, evolution (a la Seth Godin’s Survival is Not Enough) is at work, not “strategic planning” by know-it-alls with certain kinds of business or academic training.
• The “best technologies” don’t automatically win, perhaps because technology depends on how you apply it. Specific teams do well applying technologies. But more importantly the quality of relationships determines the length of client retention. Quality relationships don’t “scale.” Which means slower growth in organizations that are cobbled together out of hundreds of projects and relationships, but such organizations are robust, if not anti-fragile.

It was only after having the good fortune of reading Nassim Nicholas Taleb’s important new book, Antifragile: Things That Gain from Disorder that I realized how much my thinking on this particular issue resembles the way Taleb frames his arguments about financial markets, politics, lifestyle, and just about any domain that can apply the concept of the “convexity of risk” (which is to say, just about any domain).

(The good fortune was only doubled in that I spent some of that reading time bumping along in a bus to a mountain hike on a Caribbean island; the book stayed on the bus for the part of the hike that was down a mountain river over slippery rocks, “Rambo Style.”)

So it struck me, in comparing my brief remarks with Taleb’s system of thought, that in our field we have over here on one side, grand pundits who will typically make rather monolithic predictions about what “we’d better adjust to lest we be left behind, become toast, etc.” [Taleb amusingly refers to such types as the "Soviet-Harvard elite"]; and here on the other side — the side I chose to adopt — which was to say, as individual experts, we’re really limited in our predictive powers, and moreover, neat and tidy answers tend to be given by totalitarian thinkers who don’t give us nearly enough scope to try many different routes and fail (or succeed) in our own way.

What I probably did not see was the additional ingredient Taleb injects, which is the need for individuals, organizations, and systems to expose themselves to “small stressors” to become strong. At every level, living in a “bubble” creates an increasingly fragile organism or system until it’s obliterated in spectacular fashion. That’s one potential explanation for that weird, ho-hum organization (think of a major hotel chain or marketing services agency staffed by pretty average people and older technologies) to persist and reinvent itself constantly such that it stays in business for a remarkably long time. A knee-jerk reaction could be to simply compare the superficial impression such organizations give off to the “shiny new thing” organization that should “eat its lunch,” but somehow doesn’t. Can the explanation for the persistence of the dull organization come down to its exposure over time to trial-and-error, “many small stressors,” such that it has learned and adapted thousands of times in thousands of ways, without ever crowing about it?

In the “fragilista” lexicon of Silicon Valley and university-trained business strategists, “scale” is the answer to everything. In the language of advocates of antifragility, a failure to “scale” isn’t a shortcoming — it might in fact be an important way to inoculate a whole economic system against profound shocks. Certainly, a “winner-take-all” economy (in any realm) will pay handsomely for the winners — and is pretty good for a few related organizations that benefit from their success, certain kinds of journalists and experts who love a good “winner take all” narrative almost as much as a “collapse story.” But is it good for everyone?

Remember, Black Swan events aren’t common and by definition, you’ve never seen one of its exact type before. So “I’ve never seen a black swan” (absence of evidence) is not the same as “there are no black swans” (evidence of absence). What seems to be scalable, impressive, compelling and obviously superior to mere common-sense practices is often “proven” to be so — as Taleb evocatively points out, much like the turkey can show a beautiful Powerpoint slide of “the life of turkeys is incredibly safe, with increasing levels of statistical confidence”… each day until Thanksgiving, when the lives of turkeys become markedly less safe.

Fragility is a problem in any system or industry, and it is a problem that can be confronted and addressed. While most of us can’t affect how an entire economy works, we can assess the relevance of Taleb’s concepts to our personal finance or individual company strategies, for example. Online marketing and new types of digital companies are particularly susceptible to neomania, another Taleb bugaboo. Chasing fads and thin strategies that may have a history of all of several weeks isn’t “foresight,” it’s in fact out of step with what most sane individuals and organizations have done to govern their affairs over the past four millenia or so. What’s so wrong with combining elements of the old, tried and true, with the elements of the “new” that seem like they might have potential? Why are some pundits encouraging otherwise sane individuals and organizations to toss out almost everything that worked for them in the past and to put too many of their chips on the latest trendy tactics?

To quote Taleb at length:

“…Statistically, the ‘young’ do almost nothing. This mistake has been made by many people, but most recently I saw an angry ‘futuristic’ consultant who accuses people who don’t jump into technology of ‘thinking old’ (he is actually older than I am and, like most technomaniacs I know, looks sickly and pear-shaped and has an undefined transition between his jaw and neck). I didn’t understand why one would be acting particularly ‘old’ by loving things historical. … So by loving the classics (‘older’) I would be acting ‘older’ than if I were interested in the ‘younger’ medieval themes. This is a mistake similar to believing that one would turn into a cow by eating cow meat.” (Antifragile, p. 320.)

Online marketers and businesses looking to set priorities in customer acquisition strategies are even more fragile, typically, than they imagine. The practice of slavishly watching and relying on one’s organic rank for a single keyword, for example, would seem imprudent in the extreme, even by the measurement standards of the most “wild west” of financial-industry speculators. But in our world, that’s not uncommon at all! Imagine a bank that derived 100% of its income from loans to a single customer, at an increasingly high interest rate. That wouldn’t be a business… it would be a “Black Swan event” waiting to happen.

And consider the painful tendency of those in the SEO world to conflate one thing (the thing they obsess about) with a much more important thing; namely, their apparent tendency to equate “SEO” with “business success” in a 1:1 manner, with the latter being fully a function of the former. Anyone who actually runs a business knows that the petty obsessions of the SEO community do not map directly to business strategy and/or profitability; to cite Taleb’s archetypal Fat Tony, “they are not the same ting.” (Or as Fat Tony had it, the war about a country called Kuwait was an important development, and the price of oil definitely matters to traders. It’s just that you needed to have enough perspective to know how much the latter was a function of the former. As it turned out, they were not the same ‘ting’.) As Taleb himself adds, when considering the interrelationships among diet and lifestyle, cholesterol blood values, computed risks of cardiovascular events and death, and actual cardiovascular events and death, there isn’t anything particularly wrong with considering cholesterol blood values as a potentially useful marker. At some point in the medical literature, though, the interrelationships are assumed, and the “actual cardiovascular events and death” sort of drop out of the conversation, making it convenient to hawk potentially harmful drugs to lower borderline “blood values” in otherwise healthy patients, as if these ‘scorecard’ readings were untrammelled goods in and of themselves.

Our sector is full of fragilities that are seen as commonplace by those practicing in it. You don’t have to follow the herd, though.

More on all this shortly.

In the landing page optimization world, only a small elite reach peak performance. Those are the multitalented practitioners who have the good fortune of being fueled with data from a decade’s worth of testing, and who excel above the pack at sussing out the patterns that matter. From there, they relentlessly follow that logic in every test. (Beyond that, there is the potential for lucky accidents that lead to successful genetic mutations.)

No one has ever died from a mediocre landing page or text ad. But to borrow from Seth Godin via Charles Darwin: in business, mere survival is not enough. Let’s live larger.

Testing tiny text ads for superior performance is similar in some ways to testing landing pages. Anyone can play. A small elite “test to win.”

So, if I may borrow some terminology from Tim Ash: there are some things you absolutely must not do here (“Seven Deadly Sins”). But let’s flip that on its head: pursue these “Seven Heavenly Virtues” of paid search ad copy:

1. Focus on your core message. Simple ideas work. But it’s easy to get carried away. Product description, shipping offer, benefit, testimonial, call to action. If you have a file of ad elements that work, it’s tempting to try them all at once. But your ad will confuse people if you cram in four ideas where two will do. That must be because of the nature of the search and navigation process. The main site for persuasion is the website, so relax! Think about perfecting two ideas or elements in your ad, and if you’re good, fold in the third.
2. Love and honor your winning headline. In mature accounts, you should have reached the point where your headlines are optimized. Don’t go ripping into ad groups with a huge batch of new ads that disrespect the exalted status of your winning headline. Try new ones, but keep multiple versions of your winning-headline ads running so you don’t throw money out the door testing the upstarts.
3. Real words, please. Maybe if you’re a penny-pinching classified ad buyer renting out apartments in the local newspaper, you can get away with “2 br kit w/o deck” and its ilk. But as soon as you abbreviate, you’re speaking Greek to searchers. It just doesn’t look crisp. You’ll find the odd ad that can tolerate that compromise, but it’s likely the exception that proves the rule. So by all means, completely rewrite an ad to remove one exciting element if you need to, in order to avoid saying “brkv!” Searchers don’t know what “brkv” is.
4. Test significant contrasts in ads initially, not trivial ones. A/B testing is better than “no testing,” but it’s mediocre in the wrong hands. If you sequentially test tiny differences in your ads in a series of A/B tests until the cows come home, will you someday achieve direct response nirvana? Doubtful. Consider testing big differences in your ads in A/B or A/B/C tests. Later, consider testing more (eight plus) ad variations at once, as you get closer to perfection. Multivariate testing isn’t always possible for volume reasons, and is for advanced practitioners only. But it’s sweet when you have the volume and when you know how to make it work.
5. Use sensible date ranges when analyzing performance. There is too much seasonality in most businesses to let Valentine’s Day tussle with March Madness, or to let the August lull face off against Halloween, to say nothing of the more obvious aberrations that occur in December buying patterns. Pick smarter date ranges for comparisons, and make sure the impressions allotted to the ads in the test were roughly even.
6. Create “evergreen” ads and record-winning principles. In fast-moving industries, it’s tempting to change offers (shipping, new promotions, new products) frequently, and sometimes this is done by overwriting all the ads in your ad groups frequently. Of course, this is a must for many businesses that rely on promotions and seasonality. But the marketer’s dream is to build up a file of learnings, and you won’t get that by running a jumble of ads for a few days at a time, creating too few conversions associated with any ad creative to build up any momentum for analysis. The political and strategic challenge in any fast-moving account is to up the overall proportion of ads and ad principles that don’tchange, as a percentage of all the ads running in the account. Fast-moving promotions will at least then be able to use versions of well-tested copy.
7. Look at search-only CTR data. When you have both the search network and placement targeting enabled for the same ad group, your CTRs on ad creatives are typically reported in the aggregate. You need to unearth the “real” CTR data for ad creatives based on search-only response, and make decisions on that basis. Unfortunately the blended CTR data will look all squished together (0.71 percent, 0.68 percent, 0.68 percent, etc.), as content targeting creates a high number of impressions and a low CTR on nearly all ad creative, swamping your search-only data. Hopefully all of the search ad platforms will make this search-only reporting capability more seamless so you can do it in one click, without being forced to segregate content-only campaigns in every case.

If you fail to pay attention to these sometimes-esoteric factors, it’s just possible that you may make a living and achieve happy mediocrity — but you’ll surely never reach AdWords Heaven.

–

This timeless advice originally appeared at ClickZ on April 23, 2010. Reprinted by permission.

Seriously, by this time, wouldn’t you think they could easily filter out this type of thing?

Worse: as a user, I’m not seeing these ads because I’m doing a search for this. I”m getting it as a contextual ad after clicking on a featured news story. Yahoo populates the “search box” with terms from the headlines, then serves up three ads offering discounts on a dead woman’s name. In eBay’s case, they actually don’t appear to want to sell the corpse: it’s Death in general that is marked down. It’s like Eleanor Roosevelt said: small minds discuss people, average minds discuss events, but great minds pay $.04 if someone clicks on their ad about Death.

Yahoo has seemingly used up about 27 of its 9 lives at this point.

I know you can’t go cold turkey. But is it time to cut down on your use of negative keywords?

It might be. Telltale signs include Pop-Tarts crumbs scattered around your desk at 3 a.m., bags under your eyes, shrinking overall account volume, and a propensity to bore people at parties with material like “I still can’t believe they matched ‘Michael Bublé tickets cheap’ to ‘Costco price for Scrubbing Bubbles’!”

Worse, our industry may be feeding your addiction! I’m here to help.

As books and conference programs have become more “advanced,” we’re prone to neglecting solid strategies in favor of mastering specialized pieces of the puzzle. Today, instead of talking about what keywords are best, and what to bid on them, we can spend endless hours talking about various approaches to excluding search queries. How did that happen?

This stuff is complicated enough even when you dial back some of the ambition. Even if we make it our new mandate to dumb things down slightly, there’s no chance that the path to success is going to be easy.

Negative keywords are important, of course. They provide us more scope for elaborate tricks in our attempts to target creatively. But increasingly they’re being emphasized for the wrong reasons.

2 Good Reasons to Employ Keyword Exclusions

1. Broader match types offer convenience and reach, but mean more inappropriate matches. If you take the universe of user queries that show ads against the broad- and phrase-matched keywords in your account, and filter out 30 percent or so of those that won’t convert to a desired outcome, your ROI on that group of keywords goes up accordingly. By tapping broader match types and then sculpting them further with negatives, you enjoy significantly better returns. It’s no less important than testing landing pages to achieve 30 percent better conversion rates.
2. “Forcing” ad groups to behave as intended. This might be called ad group “quarantining.” With broader match types comes the propensity of ad groups to “steal” impressions away from “correct” ad groups. It goes something like this: you sell Johnson outboard motors and have a broad-matched phrase for “buy Johnson outboard online” among many keywords (including tighter match types) in your ad group. Unfortunately, the broad-matching technology itself doesn’t give you enough precision to keep all your ad groups distinct, so despite you wanting the users typing “where do I find zippyfast outboard motors online” to see the ad from the zippyfast ad group, the presence of the words buy, outboard, and online in the Johnson-related phrase could be enough to broad-match it to the zippyfast-related query. This now leads to worse performance in any quality scores, CTRs, ROI, etc., to say nothing of analytical confusion. The solution is to include negative keywords for all the other brand names in your Johnson group, and to do the same thing for all of your other branded ad groups. Argh! A lot of work. But depending on how you build your account, if you want to control ad delivery and ad group performance, there’s little choice.

3 Bad Reasons to Overuse Negative Keywords

1. Looking busy or following “best practices.” If you’re new to the industry, there’s nothing more exhilarating than leapfrogging the last generation of “less advanced” marketers. Skipping steps in achieving high ROI and client satisfaction and going straight to the “advanced class” of tediously negativing keywords might seem like impressive work. But who’s paying?
2. Jumping to conclusions about what is or isn’t an appropriate match. Make sure it isn’t just a fear of being “wrong” that drives you to seal your account against perceived threats. Real data will often surprise you. Intent that is “not quite right” may still convert decently.
3. Acceptance of the broad-match default, and overzealousness about “thorough” ad group build-out in long-tail ad groups. Tail keywords in tail groups will act like ticking time bombs that attract unwanted clicks if you throw them in quickly and broad match them. Bid broad matches down and beware of them generally.

Group or Campaign Level?

Negative keywords can be applied at the campaign or ad group level. We should do more to avoid too much campaign-level negative matching in general. While doing it at the campaign level is less work than custom negativing at the ad group level, campaign-level exclusions can unduly shrink the total keyword inventory available to us. By keeping more of the control at the ad group level, if we build certain groups quite restrictively, nothing stops us from building other groups with separate ideas later on, with less restrictive targeting. That approach won’t work if you’ve hamstrung the campaign with an excess of campaign-wide negatives.

What’s Being Excluded?

We often have specific intent and objectives in mind when we enter certain matches into our lists. But the lists can become so long it’s not uncommon that we run across negatives that seem likely to block searches we actually want to show up for. We’re then left to wonder if it was an error, or whether that happened for a reason: was it sound logic, a data-driven move, or an ad group quarantine strategy? (Hat tip, Matt Van Wagner, for this question.) That puzzlement can dog you even if it was you that built the ad group! The problem is compounded if multiple account managers have been in and out of there over time.

While one ambitious solution might be to create a tool or a protocol to keep one apprised of the “reason for exclusion,” a short-term fix is to reduce one’s reliance on negative keywords overall.

Building Ad Groups on Solid Foundations

If you’re seeking manageability and real-world financial outcomes rather than podium-primping perfectionism, why not reduce the number of situations that force you into advanced negative keyword strategies in the first place? One way to trigger fewer bad matches is to mandate an uncomfortably tight initial ad group build-out ritual, especially for the long-tail stuff. Stop building ad groups that are broken from the get-go.

Relying on simpler ad groups based on the most powerful match types and fewer dumps of broad-matched long-tail keywords may be a good start. Another key would be to understand sensible initial bid levels as they relate to broader match types (even with longer phrases), so the Wild West of semantic matching isn’t triggered while your back is turned.

When it comes to the inevitable need for some exclusions, there’s a way to rein in that potentially endless workload: pare down the reasons to use them to one main reason. Don’t use them until something happens that forces you to do so: namely, real-world “bad intent” queries cropping up in your search query reports in sufficient volume to matter.

The utmost in simplicity in negative keyword strategy would be to wait until your search query report shows a significant number of inappropriate queries – say, 10 to 20 on individual keywords or over 100 on several like phrases mentally grouped together – judging by performance criteria such as AdWords conversions. Want the simplest possible way to do it? Pick your date range in the AdWords interface, sort by clicks, and view the entire account’s keywords at the account level, choosing the Keyword Details > SEARCH TERMS > All report.

This isn’t a bad way to do it, but it can lead to a long, gentle drain, as thousands of under-the-radar queries numbering fewer than 10 or even five don’t convert. There’s also the problem of ad group integrity. Many advertisers will want to box out queries that should be showing against ads in another group. (But much more intergroup leakage takes place when ad groups and bid levels are shabbily structured in the first place.)

If the “slow drain” problem is predictable based on your current approach to building ad groups, it’s best to stop building ad groups that way in the first place – unless they’re significant enough in dollar terms to warrant close attention. The simplest strategy is: (a) only negativing when proven bad matches are documented in the search query report works best when married with (b) the initial strategy of building tight ad groups that won’t result in as many slow drains.

Are You Just Plain Negative-Happy?

Speaking of all the seemingly-wrong search intent affecting your campaign performance, you may find that the aggregate financial performance of a set of your bad matches is bad in your suspicious mind only. We see a lot of non-converting matches for online retailers emanating from localized queries that seem to indicate a strong desire for a brick-and-mortar experience. In the aggregate, at the right bid level, it often turns out that broad match is doing exactly what it’s supposed to do – showing ads against moderately relevant queries using predictive technology. If you’re hitting your desired metrics, relax a bit. There’s often no way to understand why Bill from Austin bought from you but Henrietta from Columbus chose not to. Combining these with all similar “semi-relevant” queries puts an appropriate price on those clicks on the whole, giving you profit on the whole. And you likely have no good way of “cherry-picking” by knowing enough in advance to somehow negative out Henrietta, even if attribution were perfect.

You’ll wear yourself out and make serious errors if you keep chasing the chimerical notion of 100 percent precision in targeting.

Don’t get me wrong: I’m a fan of using negative keywords to improve the relevance of your ads to the average user, and to improve campaign ROI. The majority of advertisers have trouble using them correctly, though, which is unsurprising given the complexity of their operation, the complexity of user behavior and account structure, and the resources at hand.

Too many others seem to use them as a fix for poor strategy and bad ad group execution.

Down the road, it may become difficult to “de-negative” an “over-negatived” account when you don’t know where to start, which can lead to anxiety around subpar volume. This can lead to overbidding on your remaining keyword inventory or experimenting with unfamiliar channels that have a significantly lower ROI than even the marginal parts of your search marketing campaigns.

That’s how it starts. A couple of exact match negative keywords at your desk one morning. Before you know it, after the cocktail reception at SES, it’s 50 broad-matched negatives at the campaign level. Soon you’ve hit rock bottom, wondering where your next click is coming from. You find yourself paying sub-dime CPMs for clicks from unfamiliar toolbars, and hanging out in the pop-under ghetto.

If this is a problem for you, I hope this column has been a step toward breaking the cycle of addiction to negative keywords. But the first step is admitting you have a problem.

Ahem. Over to you.

–

This column originally appeared at ClickZ on June 15, 2012. Reprinted by permission.