Optimization Innovations On The Google Display Network: Q & A With ScienceOp’s Dr. Bryan Minor

I was wondering the other day — how big of a player is the Google Display Network bought through Adwords with regards to the digital marketing industry? I say bought through AdWords because Google sells their GDN inventory in other ways including direct sales, their DoubleClick ad exchange platform, YouTube, etc. I’m just interested to know how big this channel is because most search engine marketers like myself only buy it via AdWords in the first place.

Let’s do some math.

I did some digging around eMarketer and found that search engine marketing comes in at around 47% of the ad revenues for the entire U.S. digital marketing industry. I’ve also seen the number floated around that SEO represents 10% or even less of SEM, so conservatively, 90% is paid search. Also, Google represents around 83% of U.S. search advertising revenue. It’s fuzzy math, but it seems possible that Google AdWords is about 35% of the whole pie. That’s a nice slice of a multi-billion dollar industry.

To answer my initial question, I need to know what percentage of that is paid search and what percentage is the GDN in AdWords? If you have the actual public data, please post below in the comments section, but I’ve heard that it could be somewhere between 20-40%.

That means that the GDN channel bought through AdWords just might be 7-14% of the entire U.S. digital advertising industry…or more! And it’s growing faster than search as more and more sites implement the AdSense code to be part of this very lucrative network.

It is kind of bewildering that not many technology solutions have arisen to focus on this lucrative channel. For the most part, I think that we are trained to optimize content and placement campaigns (a.k.a. automated placements and managed placements) the same way that we optimize search.

Basically, we bid low or pause poor performing sites and promote the ones that show a better return—just like we would keywords in a search targeted campaign. We may tinker around with our keywords and ad groups in our content campaigns, but it’s still just more pausing and promoting, right?

So, where’s the tool innovation to help search marketers better optimize this channel?

I recently met Dr. Bryan Minor, CEO/President of ScienceOps, whose company is tackling this very issue. His online bio begins with: “Dr. Minor is an expert in computational physics, parallel and grid computing, nuclear physics, radiation transport, and numerical methods.” With the home page quote of “infinitely intelligent algorithms” and a product team that includes former NASA scientists (yea, I said NASA), Bryan hopes that his platform, AdMetrica®, might just be the solution that search marketers are looking for.

Q: What is your background and how did you get into display advertising?

Bryan: For almost ten years now, ScienceOps has been in the business of solving extremely difficult problems through algorithms. Since 2001, ScienceOps’ only method of advertising has been Google AdWords, and success in this marketing channel was required for our existence. In February 2007, we discovered that almost 90% of our leads came from the Google Display Network (GDN), then called the Google Content Network.

This discovery started our focus on optimizing GDN advertising performance and resulting AdMetrica® technology now available to all advertisers. In 2007, it became clear that the advertising opportunity in display was huge, and that display needed to be treated much differently than search. We approached the optimization of display advertising as we approached problem-solving for clients in realms such a genetics, space travel, and nuclear  physics – by applying our knowledge of mathematics and science.

Q: What problem does your company hope to solve?

Bryan: Generally speaking, we’re solving the basic problem of understanding the GDN. Not just the mechanics of its algorithms, but the important implications related to controlling costs and gaining meaningful and sustained conversion volumes. High costs and low returns have clearly been a problem with GDN advertising for many advertisers. We’re working in collaboration with Google – within AdWords – to tackle those problems head-on, and, frankly, we’re winning.

Q: How are you doing that?

Bryan: It’s a multi-dimensional solution centered around AdMetrica’s suite of algorithms. It starts with continually monitoring and synchronizing with Google, and adapting to their changing algorithms. That information from Google is the basis for all the decisions made by AdMetrica. But it also gives us a view, at the human level, as to what’s happening and why. For instance, it allows us to see the various pages where ads are being placed; we’re able to track the exact URL’s that are getting the most conversions.

On another level, AdMetrica manages each of the major performance variables: keywords, bids, placements, and budgets. Keyword lists, for example, they’re automatically sampled, tested, and then rewarded according to how they’re performing. That alone is a monumental task, beyond the scope of humans. It’s the pivotal piece in generating conversion volume. It also serves to lower CPA. Of course, suppressing CPA is the other half the performance equation. Solving for lower CPA’s and higher conversions in an unending challenge we enjoy.

Q: What kind of results have you seen?

Bryan: For the right type of advertiser, the results are pretty extraordinary. We expect to see conversion increases ranging from 30 to 120% during the first month, while at the same time seeing double digit drops in CPA. That might seem counter-intuitive until you understand how Google rewards performance. Having said that, though, this technology is not for everyone. The type of advertisers best suited for this technology have a broadly appealing offer to a large geographic market. It’s especially powerful for companies like Meredith Media that are already doing well with GDN but want to do a lot better.

Sure, we can optimize for niche players with less reach, but large advertisers are able to produce the data flow needed for AdMetrica to sustain its required learnings. Other characteristics we look for to gauge the likelihood of success are such things as having a strong brand and that have the ability to produce a landing page with a low conversion threshold. Fortunately, we’ve been doing this long enough to have a pretty good idea beforehand who will perform and who won’t.

Q: What were some assumptions that your team had going into development of your tool? What did you find to be true?

Bryan: Well, we’re scientists so that question goes to hypothesis. We assumed that advertising opportunities on the GDN would grow significantly. We saw early evidence indicating that Google “content” (before it was the named the GCN, before it even had a name) could produce a larger volume of potential customers than search. These assumptions are proving to be correct but it’s an ongoing study. The problem then quickly turned to accessing GDN customers. That’s an incredibly difficult problem; in some ways more difficult than astrophysics.

To start, we approached the problem differently. Early on, we understood the importance of having the correct keyword lists and the difficulty this non-linear problem presented. The sheer size of the keyword search space is daunting, but it’s also exciting considering the enormous size of the advertising opportunity. With this technology, we ensure that our client’s keyword lists continue to be explored and refined. Everyday we gain deeper understanding of this key issue, and benefits of all our clients.

Of course, we also assumed that it was important to identify the URL locations upon which an ad is being placed. That was central to the development of AdMetrica. It allows us to see a rank-ordered listing of an ad’s best performing sites. Being able to then monitor and optimize for them has been invaluable for both volume and costs.

Q: What have been the biggest challenges to your development?

Bryan: The biggest challenges probably stem from the fact that nobody’s ever done this before. It seems that nearly every day, certainly every week, we’re uncovering new findings that reveal new opportunities for evolving performance outputs. And these are not small findings. Like any complex issue we’ve explored, the deeper we delve into it, the more variables we see, and the more connections we see between them. The fact that the variables are always moving is one of the biggest challenges. The good news is that our scientists are well versed in dealing with dynamic variables in highly complex environments.

Q: What does the client interface look like?

Bryan: For those who know AdWords, it’s very familiar. I’m kind of joking, actually, because AdWords is the interface. This means that nothing about the advertiser’s workflow changes. It would be hard to create a better interface than AdWords anyway, so this approach makes it simple for the advertiser. For them, it’s just a matter of watching performance numbers change; there’s no new interface to learn.

Q: How do you see your company evolving in the next 3-5 years?

Bryan: Wow. That’s a loaded question. We’ll continue to be a scientific organization. We’ll always be rigorous in product testing because without accuracy you have nothing. Beyond that, we’re mapping the GDN. We’re striving to better understand of how the GDN affects behaviors of all the other ad channels: SEM, SEO, and even typing the URL directly into the browser.

We’re seeing the holistic nature of the online advertising environment as a whole so that’s a rich learning source, but only as a means to evolve our knowledge of display advertising. We’ll remain dedicated to display.That’s simply because there’s more value in display – especially in the GDN – to the advertisers. Display is a space that’s all but unexplored; it requires our full and deepest attention.