Author Archive


First Impressions Count, But Ad Impressions May Not

Written by Andrew Edwards. Posted in Paid Search

Artist impression of a GPS satellite model viewed from the gallery sweeping elevated walkway. C. 2013 Science Museum Universal Design StudioRecently, a customer suffered a costly setback in a CPM campaign — and I suspect they’re not the only ones with the problem. In fact, I think there’s ample evidence to suggest the CPM and PPC world is rife with opacity and strange artifacts.

Here’s what happened: the advertiser runs a professional services firm and wanted their ads to appear at certain sites where they believed customers might be found. That sounds simple enough. They contacted a relatively small ad network that claimed to specialize. The CPM program (which charges per impression, not per click-through) ran for several months and cost nearly twenty thousand dollars. It was canceled due to non-performance.

Why did this happen?

Analytics Review

A review of the firm’s Google Analytics revealed the traffic sources from the campaign and the amount of impressions from each venue. It turned out that over twenty per cent of the impressions delivered came from a site owned by the ad network itself. This seemed an anomaly. Upon further investigation, the site proved to be a page full of videos for a variety of brands. It had no other content besides the videos and there seemed little reason for anyone to actually visit the site. So where did all the impressions come from?

It’s really anyone’s guess. But one might conclude this URL was used simply to generate impressions with no actual visitors or at least, no visitors relevant to the campaign. When the facts were presented to the network, they credited the customer for all the impressions from that site.

Score a win for analytics!

But what if you’re running multiple campaigns and don’t have time to run down each and every bad batch of impressions? How would you know where your impressions were coming from, whether they were actually visible, and whether the site was of any real quality relative to your goals?

comScore, the well known media measurement firm, may have some answers. While they recently came out with a product called Validated Media Essentials (vME) designed to help media sellers increase advertising inventory and revenue, they also have a product called Validated Campaign Essentials (vCE) that is designed to help buyers fight opacity in networked advertising.

Here are some of the things it can do:

  • Enables in-flight campaign management and optimization.
  • Evaluates audience delivery, viewability, brand safety, geographic delivery, engagement and non-human traffic.
  • Reports data by publisher, placement and creative.

The net result of these kinds of insights is that the buyer can increase campaign effectiveness and, just as importantly, they play an important role in decreasing wasted impressions (and dollars).

In the newest version of vCE, you can gain access to important data based on an accounting of impressions delivered across a variety of dimensions, such as ads delivered in-view, in the right geography, in a brand safe environment and absent of non-human traffic. It also evaluates the degree to which validated impressions reached the campaign target audience.

Why does this matter?


Geographic data is key because ad networks use a wide variety of sites to deliver impressions. But if you have an interest in generating traffic within a specific region, your ad network won’t necessarily target geography, or at least not with sufficient accuracy. Tools like vCE 2.0 will help you understand where the traffic is coming from. This is accomplished by comScore because they rely in part on a panel of known users. By extrapolating data from these users (there are over a million in the US), they can give you a good idea of what the geographic profile looks like for your impressions.


Viewability is also a key metric, yet your typical analytics tool will not provide any insight into this. The term refers to whether or not your ad, even if served, actually showed up in the screen seen by the user. Think of it this way: many pages are longer than a single screen, and often you have to scroll to see the entire page. What if your ad was served, but was too far down the page? And what if the user left the page without ever seeing it? vCE lets you know whether or not the ad was actually capable of being seen, or if it was instead hidden “below the fold” (to use an old newspaper phrase) and never actually viewed by a person. You should not have to pay for ads never capable of being seen.

Brand Safety

Brand safety is often overlooked, but it’s the equivalent of insisting that your billboard for milk does not appear on top of a slaughterhouse. How can you know that the sites where your ad is seen don’t in fact make your ad reflect poorly on your company? What is the quality of that site? Does it have lots of negative comments in its forums? Is it badly formatted, possessed only of thin, irrelevant content? Does it seem to present an unsavory image for your brand? Make sure your ads are appearing only in places that make you seem both relevant and elegant.

Non-human Traffic

Non-human traffic is also a major problem, as pointed out in my example above. There’s no single reason why a particular ad network might generate impressions to non-human traffic, and we’re not here to determine whether this is ever done on purpose to inflate impressions. But it’s not an insignificant number of impressions in many cases. And without a way to track it down, you’ll pay for ads that somehow got served, but in a manner not associated with human activity. Unfortunately, bots today do not buy things. Perhaps one day they shall (as we enter the Age of Drones) but right now they are not very good prospects. There is no reason to pay for these impressions.

Piercing the Veil

Many buyers believe, mistakenly, that CPM campaigns are a set-it-and-forget it proposition, and that they are getting a clean deal from their ad network. But there are too many variables involved, and ad networks sometimes prefer opacity to accountability.

You may not need a specialized tool like vCE — as indicated above, you can certainly learn much just from reviewing your standard analytics with care. But vCE and tools like it go beyond what analytics alone can provide. Especially if you’re managing high-volume campaigns, it would be a mistake to rely simply on reporting from the ad network. Self-reporting is too one-sided. You need a way to arbitrate your buy. Using third party tools to measure ad buys is just as important as using third party analytics to understand traffic patterns.

Don’t rely on first impressions when it comes to ad impressions. Dig deeper and you’ll uncover more value.



Digital Analytics in an Age of Spying

Written by Andrew Edwards. Posted in Privacy

snooping Q: How many NSA spies does it take to screw in a lightbulb?

A: At least a hundred thousand; one to twist the bulb and the rest to monitor everything that person and everyone they know have done online since 2004.

Gone are the days when the worst a privacy zealot could shout (petulantly) about was the fact that ad targeting was like “Microsoft putting a billboard on your front lawn,” to paraphrase the soon-to-retire Walt Mossberg of the Wall Street Journal. It was easy to put the kibosh on this kind of argument.

Privacy, yes. Invisibility, no.

The refutation went something like this:

When you’re cruising the good old Information Superhighway (the one “Al Gore invented”), you’re traveling outside the confines of your own front yard. You’re going to places owned by other people or other businesses. You’re pulling stuff off their servers. No one forced you to go there. You went there on your own. And just as if you were visiting a shop and the shopkeeper got to know whether you liked Bufferin or Excedrin, in the more sophisticated digital environment, the behavior you exhibited began to create a “persona” that reflected your apparent tastes and affinities. So it should come as no surprise at all when marketers, who think they have something you might be interested in, start putting ads before you (in your browser) that might entice you to buy. The horror!

Dark Lords

Of course, the most advanced digital marketers have known all along that Bed Bath & Beyond wasn’t the only outfit interested in what you did in the privacy of your own home (no knock on BB&B). But for nearly everyone else, the Age of Innocence was shattered when a certain Snowden individual (perhaps it’s fitting he shares a name with a character in Catch-22) revealed that the NSA was tapping into the internet to penetrate the privacy of pretty much everyone, pretty much all the time.

The fact he was a low-level contractor (one of thousands) working at a private company — and that he had enough security clearance to be able to type in random stuff on his computer and come up with random data about pretty much anyone — made it only more chilling for those who understand the mechanisms of police states. It’s bad enough that someone very qualified and very scrupulous might be scanning your emails and phone calls. It’s far worse to know that any of a horde of back-benchers might decide to mine your data and do who-knows-what with it.

This news can have few fans outside of the security state. They will claim it’s all done to thwart terrorism. Ben Franklin would have disagreed: “Those who give up liberty in pursuit of security deserve neither.” Publishers can’t stand it, because it impinges on their freedom of speech. Internet Service Providers hate it because it makes them into compliant conduits of private data they had previously promised never to reveal to anyone.

The Business of America Used to be Business, But Now It’s Security

Now, apparently, Cisco hates it too. They just had a no good, terrible, very bad quarter because of a serious drop-off in orders from abroad, according to an article in Cisco had projected to grow its overseas orders by 6 percent. Instead, in Brazil for instance, they dropped 25 percent. Cisco has made statements that seem to attribute this either to fear amongst foreign corporations that American companies are too much of a security risk for them to buy products and services from (because, ostensibly, they’re not sure if the American pipes won’t lead right to Langley), or simply in retaliation, based on the sheer anger at having been swindled about America’s commitment to freedom.

Either way, it’s a bad scenario, thanks very much. American companies are getting hurt. Does this mean the terrorists are winning? Isn’t this exactly what they’d have wanted?

Hating on Digital

Digital marketers need to be concerned, as well.

In a recent Piedmont marketing class, the professor showed young marketers how they could use Google, YouTube and Facebook to help market businesses. He might have been surprised when his students told him they really didn’t like the idea, and that social media generally seemed at best a bother and at worst kind of creepy.

Clueless newbies in a backwater? Or is there a backlash in the works?

It could be either of the above.

What’s more important is that if there is a backlash, it’s not just because people may be tiring of tweeting. It’s that the underlying perception of digital analytics has, amongst the polity, taken a nosedive. Even as marketers begin to find Big Data and Hyper Data and microtargeting and campaign attribution and sophisticated modeling and even real actionability now within their grasp, they are coming up against the fear of tracking in general.

Nevermind that there’s no connection between Zappos remembering what height of heel you bought last time and the NSA recording the conversation you had the other night with you-know-who. The perception grows that it’s all just one big bucket of nasty fish and they’d rather someone chucked it back into the drink.

It’s not fair, but in a digital world where perception is reality more than ever, how much does fairness really matter?

Marketers should take a look at what happened to Cisco. They should be worried that the revelations about spying—spying for which both Republicans and Democrats are responsible—are creating an atmosphere of mistrust not only for digital analytics, but for American companies in general.

Perhaps the NSA can set up a superfund to clean up this toxic mess before it swallows the whole neighborhood.


Can Analytics Help Save Obamacare?

Written by Andrew Edwards. Posted in Analytics

The crash-on-takeoff of the Affordable Care Act website is already one of the most notorious flameouts in the history of digital enterprise.

Putting any merits or demerits of the ACA aside, we must grapple with the fact that as I’m writing this (one month after its launch), the digital platform upon which the ACA is perhaps too heavily reliant has been a near-total failure.

Many will be tempted to suggest this is because it’s a government effort and that government never gets anything right; it’s a sink of waste and foolishness and this proves how it ought to have been left to “the private sector” to construct and manage. But this is not necessarily the case, not least because the act was largely “privatized” in the first place, to placate a conservative congress.

Would the ACA be a much simpler proposition in a single-payer environment? No doubt. One of the the key challenges seems to be the coordination of numerous private databases in the service of what is essentially an e-commerce site. It really is not much different than a site that sells shoes, with special discounts for those who qualify. On the ACA site, you “shop for insurance.” Perhaps they ought to have given the whole project over to Amazon and had Amazon put a “healthcare” tab on the home page.

The difficulty with the the ACA web sites, apparently, is that many who seek to enroll cannot get past the authentication page, and cannot therefore apply for insurance.

To address the challenge, the White House has promised a “tech surge,” which seems to be taking shape in the form of a cry for help to Silicon Valley. It’s the first thing we’ve heard so far that seems hopeful and we wish them all the best.

Analytics or Continued Failure

The “fixers,” whomever they turn out to be, will need to pay close attention to analytics. At first, it will be all about error messages and fixing infrastructure. But once they get it to perform minimally, it will be, because of its early failure, just that much more important that the site delivers on conversion. Those in charge will need to rapidly identify friction in the system and make it more than a default way of solving a problem.

Why Digital?

We come quickly to the question of whether digital is the right way to have tried to solve the health care morass.

The prevailing wisdom is that digital is the savior; that it cuts friction, that it is swift and simple and creates a permanent solution. However, it is entirely possible that a system of 1-800 numbers and paper applications might have served as well or better than an online solution (of course, they will have needed the same massive database on the back-end, but that is another matter). The ACA may quite unexpectedly prove a watershed in our belief that digital can solve societal problems and instead prove that it really cannot–at least not without much, much more care and devotion that would have been required by another (non-digital) system.

Analytics is What Makes Digital Different

But digital compensates in many ways that non-digital cannot hope to compete.

The way that digital triumphs comes down to its built-in accountability. Let’s not try in this brief space to describe the KPIs and reporting dashboards that the ACA site operators must put in place immediately to understand user behavior and make the site as efficient as, say, Amazon.

Let’s say instead that without the ability to:

  • understand user behavior and spot trouble and fix it quickly,
  • continually test and improve using feedback and algorithms,
  • define desired outcomes and build in measurement and optimization;

…then digital really becomes just another choice among choices, and who can say if it was the best?

We suspect that this method was not followed in the development of the current infrastructure, and the results are a very poor showing indeed.

The only way to prove that the website for the ACA not only works, but works well–and works better than another system might have worked–is to apply robust site analytics, take careful note of what the data is saying, adjust accordingly and thereby demonstrate that digital is not just an obvious choice, but an essential component of the program’s success.

About Efectyv Digital

Efectyv Digital is focused on strategy for two distinct markets: digital analytics end-users; and marketing strategy for technology companies.

Click here to learn how how we can help your business grow >