Chris Zaharias runs Sales for Omniture’s SearchCenter SEM solution, and is a seasoned sales executive whose working knowledge of search marketing dates back to the late 90’s. Chris has more than a dozen years of ecommerce, online marketing and enterprise technology experience, including domestic and international management positions at Netscape, RealNames – an early pioneer in the paid keyword space – and Efficient Frontier. Chris is an avid SEM writer and blogger and loves nothing more than knowing what’s new in search.
Eric Enge: What are the areas where search and traditional marketing are intersecting?
Chris Zaharias: Advertisers have realized that there are a lot of interactions between search and other forms of marketing. If we were to talk about this four years ago, I would have said the interactions between search in the other forms of marketing and an advertiser’s own business were immaterial, because the greatest opportunity for growth in search had nothing to do with that.
It had everything to do with building out keywords sets, making sure that you have the right ad copy, sending the traffic to the right page and doing all of the SEM 101-type of things that most advertisers did. What’s changed now is that advertisers, at least in the US, are already buying most of or all of the keywords that they should be. They’ve done a lot of things, like structuring their campaigns correctly and participating on all three or four of the major search engines, as opposed to just one or two.
Those types of organic growth opportunities have dried up. But at the same time, the overall growth in search as measured by metrics like the number of searches per searcher have also come to a virtual standstill. There has been a lot more focus on how to properly manage search campaigns and what to do to set yourself apart from the competition. The competition has figured out all of those SEM 101 types of tactics because of all the writing, the conferences and the increased focus on this particular.
So this is where it really gets interesting to try and understand search marketing as part of an overall marketing mix and to use that knowledge to get more out of search specifically.
Eric Enge: To summarize, growth was sort of brainless for a while, but that is not good enough anymore.
Chris Zaharias: Yes. I used to always say that all people needed to be successful in search in the first few years was a pulse. It really wasn’t that hard because it was such a green-field opportunity.
Eric Enge: Right, but in the current environment, where a lot of that is more mature, it’s not sufficient. Your company might get some gains out of that, but it is not differentiated.
Chris Zaharias: Right. There is also the dynamic of the search engines themselves. Because of the slowdown in growth and the continued imperatives on their part to grow their businesses, they have now started to operate their engines with an eye towards yield maximization. There are a lot of things that search engines have introduced that increase their monetization but require more complexity in campaign management.
They have also introduced more requirements for precise and holistic management of the campaigns. This is done in order to create yield management on the part of the advertiser to combat the monetization moves that the search engines are making.
Eric Enge: So you can still get a competitive edge by better managing these second generation opportunities (problems) than the next guy.
Chris Zaharias: One of the key opportunities for growth and improvement of campaigns is conversion optimization. When you look at any given sector in paid search, you will find there are up to order of magnitude differences in the conversion rates that advertisers are getting. SEM tactics being equal, the advertisers that can convert the traffic at the highest rate are going to win.
Whereas the focus for conversion rate improvements used to be just iterating your ad copy or trying to see which page converted best, there are now requirements to get much more scientific about the conversion optimization process. One area that jumps out is known as Multivariate Testing.
Multivariate Testing is testing all of the potential variations of a landing page or a session. This can include the ad copy, landing page, and all of the different pages in the conversion funnel. And on each of the pages of the site there are multiple elements that can be tested, such as the offer itself, page layout, placement and size of graphics, font size, color schemes, and much more.
Breaking down all those variables into a set of tests often yields conclusive results as to what combinations of elements best convert the traffic.
That’s the reason Omniture acquired Offermatica in late 2007. Omniture realized that the opportunity to make big strides in conversion optimization was there, but that it would require technology that marketers could manage to take control of and continually iterate through the testing process. So, one big area is conversion optimization.
That industry has been getting more refined in understanding how to manage paid search campaigns. Historically, people started out optimizing paid search to clicks to pay for their traffic. The next wave was from 2003 to 2007-2008, and it concentrated on conversion for people who are optimizing to a cost-per-order or CPA metric.
Now a new area of opportunity for greater efficiency is getting more fine-tuned in the metrics. When you look at a typical keyword campaign, you will find that high-volume keywords tend to have enough data to optimize your conversion metric.
But for the midsection and the tail of your keyword portfolio, you tend to not have enough data to optimize to conversions and you essentially have to use your best human intuition to make decisions. That keeps you from having the efficiencies you want in your campaigns.
We are seeing a lot of people use micro-conversion events, also known as leading indicator metrics.. What that might mean in a retail setting is rather than just optimizing to the conversion event, you might try to understand the ratios of events such as shopping cart fill or viewing a product detail page. You can assign values to the leading indicator metrics. You can optimize using a lot more data than if you were to just look at conversion data.
Eric Enge: That can include things like newsletter signups or contact box requests.
Chris Zaharias: There are literally hundreds of metrics, but the point is defining and assigning values to those metrics. Then you have to optimize to those metrics, which is something that very few people are doing, but something that everyone should do. It just increases the efficiencies of your search marketing campaigns that much.
As result, integrating your web analytics with your SEM Campaigns becomes very critical. If you can’t be aware of those metrics or analyze data to understand the ratios of those micro-conversion events then you can’t optimize.
The third area is this notion of trying to understand SEM within the context of your overall set of marketing initiatives and your overall business. There are consistent patterns of interaction between TV advertising and search that you can track, analyze and optimize. One example could be overlaying search traffic and conversion data against TV advertising data. It’s clear when you look at this data that the TV activity drives search activity in a meaningful way.
What that means from a campaign management standpoint is that you need to integrate those other forms of marketing into how you manage paid search. If you know you are going to be running a set of TV campaigns, for example, you may want to adjust your bid management rules in advance of the expected rise in traffic during that time.
You could say the same thing for whatever medium you are advertising in, they are all going to have an impact on search. For that matter search is going to have an impact on other forms of marketing.
Another thing to remember is that you can’t be truly efficient in the management of your search campaigns unless you take into account the direct and indirect online value that you accrue from those activities. But you also have to take into account things that might happen offline. This has been a big challenge, particularly for the SEM community, because it is comprised primarily of SEM-specific tools, vendors or agencies.
When an advertiser sees that there is a competitor that has a big offline presence that appears to dominate the paid search space for their sector, they will always ask how can they afford to consistently spend 40% more than me when I am doing everything right and taking into account all my online value that I am getting from paid search?
The answer is that the advertiser is able to understand the value that they are getting from their search campaigns in their physical stores or something along those lines.
Eric Enge: The classic scenario is that someone does a search for “women’s clothing, and then, they want to go see it, they want to touch it. They decided what they wanted to buy based on search, but then they went in the store to complete the transaction.
Chris Zaharias: Yes, the online-only retailer is always going to be at a disadvantage, because they can’t consummate the subset of transactions that are going to happen offline. Or, if you are looking at two advertisers that are both in search and have online and offline presences, the one that can instrument their campaign management to actually understand the exact relationship between a particular keyword and a particular offline transaction is going to be much more efficient than one that is limited to knowing that 20 percent of its business happens offline and therefore assumes that each search driven online transaction has a value of 1.2. That would be a very approximate way to do it.
What advertisers need to do is integrate their point of sale systems and their physical stores with their analytics and search marketing systems so offline data can be piped into their analytics and SEM environment. If they do this they can actually take into account the direct relationships between a query and the ultimate offline transaction.
Eric Enge: That sounds like a fairly complex process. Say someone comes to the website and they fish around for a little bit without leaving a name or anything like that behind, how do you create the correlation to when the transaction takes place in the store?
Chris Zaharias: It requires a very specific method of tracking. Let’s use a call center as an example. If we are talking about a phone order, what you might do is have a particular 800 number or coupon code that you show to the person that came through Google campaign X.
This way, when they call in you can correlate that back to a particular keyword or campaign just by virtue of the number they call. If you are trying to entice people with some sort of promotion, you might want to tie back that promotion to that promotion code or a particular keyword, or particular campaign, or particular engine.
All you are doing is correlating with the particular click-through that happened from paid search to make sure that you take the value into account. The main reason I came to Omniture is because I felt like the system to do this in a much more exact fashion was going to become critical. The acquisition of Visual Sciences gives us a system that integrates with the point of sale system or the call center system. Then it sends the data from that system to our SiteCatalyst infrastructure, which is the data hub for analytics.
Because Omniture SearchCenter is integrated with SiteCatalyst, we can have SearchCenter act on any source of data, including that call center offline data as it figures out how to bid on keywords.
Eric Enge: Right. Calling a custom 800 number is actually much simpler than capturing the coupon code at the point of the sale because it is fully automated. The other thing to realize is that even though you have improved your situation greatly, you are still only capturing a percentage of the transactions. You are not capturing a hundred percent of the transactions, you are just capturing a chunk of them.
Chris Zaharias: That is true. But every increment of efficiency is good to have.
Eric Enge: Absolutely, the more information the better. Even though we know that analytics is an inexact science and everybody talks about how relative measurement is king, you are still better off every time you remove a source of error.
Chris Zaharias: Right. So, in addition to taking offline call centers and what data you have into account, to understand its impact on the search marketing campaigns and react to it, there is also the continual goal of targeting to or taking advantage of all the targeting capabilities that exist in search.
What you might do is try to use that offline data to understand the impact of search in a particular geography. We have certain clients correlating conversion data back to the geo-targeting they are doing in Google to perform geo-analyses of offline conversion data. Because the response from search tends to drive offline conversions of different amounts in different geographies, that’s going to affect how a company actually implements and iterates through geo-targeting in the search engines.
Eric Enge: You might spend more on your Boston campaigns then your San Jose campaign, for example?
Chris Zaharias: Correct. That’s one of the main areas where Omniture is going with search and integrating with both SiteCatalyst and the other pieces of Omniture’s platform.
Eric Enge: People used to believe that the CPA was everything. Don’t get me wrong, it’s a great and reasonable way of measuring things, but why not use real revenue goals instead? Once that occurs, the next logical thing is to use gross profit. On each step along the way, you are eliminating a source of error.
Chris Zaharias: Exactly. I just had this conversation with someone in the credit scoring space a couple of weeks ago. They are spending $25 million plus a year on search, but the one thing that they haven’t been able to do in their campaigns is optimize something other than CPA. They obviously want to optimize to profit, which requires that they take Lifetime Value (LTV) into account.
Ideally we would take the LTV data, get it into SiteCatalyst, and then use bid rules that take into account the variable profit of each transaction that they are going get from search. Say you sign someone up and they are paying a monthly fee for credit monitoring services. Compare someone who stays with them for just a one month trial then cancels, with someone who stays with them for three months, with someone who stays with them for 12 months and buys an additional product.
They are very different values, and unless you take those different values into account, you are never going to be as efficient as you could be in your search campaigns.
Eric Enge: The people who convert as a result of one group of keywords may be more likely to be long term customers than those that convert on another group of keywords.
Chris Zaharias: Right. Before the whole macroeconomic shift, the sole focus of paid search campaigns tended to be top line growth. But now we are hearing a lot more people who don’t care if it makes them money or saves them money, as long as they increase profits. These types of discussions about increasing efficiency, regardless of where they come from, are a lot more prevalent.
You talk about how the search engines are trying to monetize their inventory. The need for analyzing site traffic data to find and implement negative keyword strategies is a lot more acute. And you could argue that an advertiser who is doing everything right should have as many negative keywords as positive keywords. Typically, however, an advertiser will have maybe one negative keyword for every five or ten positive keywords.
If you can help people who are making use of broad match, find all of the areas where the search engines are not properly matching the ad to the query, you can help them build out those negative keywords. This way you can drive a lot of efficiencies that the search engines are unlikely to show you.
Eric Enge: There are a lot of flavors of those kinds of things. It could be as simple as deciding between broad match versus phrase match.
Chris Zaharias: The key to finding the right match type strategy is to think of a broad match type as a baseball farm team. You have a lot of players in there, some of them are going to pan out and some of them aren’t. But you need that farm league system to continually grow your campaigns. It’s going to feed both the positive keywords as well as help you find all of the negative keywords that you want to avoid at all cost while using broad match.
Have you heard of this automatic match beta that’s been going on at Google? It is very emblematic of the direction that the search engines are going in. Some people are discussing whether or not bid management is dead. The real discussion that should be going on is whether or not the keyword is dead. In an ideal world, search engines would go to an advertiser and say, you are in this vertical, let us figure out all the keywords and all the traffic that’s appropriate for that vertical.
That would be great for the search engines because it would allow them to cut out the middleman and get a lot more revenue than the conversion data would merit. The reality is that’s the opposite of what advertisers should do. Advertisers should take advantage of the ability to target at the individual keyword level to find the right buying intent to go after and capture. So search engines and advertisers are increasingly at odds with each other from a strategic perspective. This is where the online business optimization platform that Omniture is bringing to bear becomes much more compelling.
Eric Enge: Right. So, for an example let us say that you are an online shoe site and you focus on selling men’s shoes. Let’s say you don’t sell slippers, so you really don’t want to be matched up with them and you never use the word “slippers” in your campaigns. But, with automatic matching, you might end up showing up for that.
Chris Zaharias: Right. Think of Automatic Matching as a beta system whereby Google looks across accounts for unspent budget and then broadens the ad-to-query matching as necessary to spend the remainder of the budget.
Eric Enge: Let’s talk a little bit more about TV driving search. Let’s say you run a Super Bowl ad. You should see a pretty good spike in your search traffic as a result, particularly if you are smart and prominently featuring your webpage address in the ad. But in addition there will be people who search on your brand name. So, what’s the best-practice for trying to measure that kind of impact?
Chris Zaharias: Well, I think that being a very unique advertising scenario, where there is an immense amount of money you are spending in an incredible short period of time, it’s definitely a bit different than just the standard ongoing TV campaigns. If you are going spend that much money for a 30- or 60-second spot, you need to be confident that you are going to be able to convert that traffic at the absolute highest rate possible.
In that specific case you would be criminally negligent if you didn’t do some multivariate testing in advance of that TV spot to make sure that it converts at a very optimal rate. The very first thing to think about would be testing to make sure that you have the right flow, the right offer and the right formation of pages that optimize that expensive traffic you are going to be paying for.
Another thing from an SEM management standpoint is to know exactly when your ads going to run. You are going to want to make sure you implement a set of management rules for your SEM campaigns that take into account what’s likely to happen that day in advance of the Super Bowl.
You might have to up your budget for this period, or you may have to impose a different set of bid rules for a period of time after it airs to take into account what is likely to be a much higher conversion rate than you typically get.
I am also going to revert back to my long standing keyword management principles as the effect of the Super Bowl advertising trails off. And that requires some specific thought and advancement in infrastructure. Let’s say you go into SearchCenter and you look on the calendar, and you impose a specific set of bid rules for this specific period of time. That could either be a period of days or a specific set of hours within the day that correspond with when your ad is going run.
Depending on the type of TV advertising that you do, you are going to have to think about different keyword sets, both positive and negative. If ever there is a time when you need to be aware of all possible misspellings of your brand name, this is it. People that are four beers into watching the Super Bowl they might recollect your brand name correctly when they see an ad or they might have variations; making sure you capture all of that is going to be critical.
Likewise, making sure that you have all the negatives set up to avoid getting traffic that you don’t want becomes even more important if you are spending about 2 million bucks for a 30- or 60-second spot. You might have specific reporting that you need to see during and after the Super Bowl ad is run.
Obviously, if you are an SEM manager at a company and all of a sudden you are going to be spending 2 million bucks on a Super Bowl ad, you can bet that the reporting requirements of your senior management around your SEM campaigns are going to be very different during that time than it has been to date. You need to be able to provide near instantaneous detailed executive reporting so that they can understand the value of that investment.
Eric Enge: Right. So, you can do differential measurements, correct? You can compare the results on the website for the 7 days following the event verses the results for 7 days previous to the event, for example.
Chris Zaharias: Right.
Eric Enge: So even though we may not have a precise number on the income that it brought in, we do know that income was 60% higher during that 7-day period, than the prior 7 days.
Chris Zaharias: Right. Every metric that you can imagine from transaction volume, to average transaction size, to geographic measurements to conversion rate is going to be looked at.
You need to be managing your SEM campaigns in a manner that takes the interplay into account, not just with offline marketing channels, but with other online channels as well. There is a value in someone who might sign up for a newsletter on your site as a result of seeing the ad or who might play a game that you develop on your site specific to that Super Bowl ad, and whose goal is engagement and brand awareness with the target demographic.
Being able to have your SEM systems integrated with your email campaign management systems becomes very important if you are going to achieve the engagement and conversion goals that you have.
Eric Enge: What if we have an ongoing TV campaign rather than a spike event?
Chris Zaharias: In that case you really need to make sure that you are understanding, capturing, and reacting to the value that you are getting from the offline campaigns. You will want to have your SEM management set up to be able to bid in the way you want to during the times those ads are running. Obviously, you will typically get spikes in traffic and changes in conversion rates relative to other times when you run those ads.
If you just let your system react to it, the system is inevitably going to react poorly because it has no reason to know that anything is about the change. It will react slowly and then efficiently to changes in traffic conversion rates.
What you should do is tell the system that you want to bid up or relax your CPA requirement in anticipation of a higher conversion rate or higher average order sizes during this period. Doing that analysis and implementing the rules to coordinate your TV campaigns with your search management is going to yield better results than just letting it ride.
Eric Enge: Right. It could be something as simple as being willing to take a lower margin because you expect a good spike.
Chris Zaharias: Yes and SEM systems act on historical data. It’s great to use historical data, but only if it’s a good representation of likely future performance, which isn’t necessarily going to be the case in TV advertising.
Eric Enge: Right, absolutely. So this sort of gets back to Omniture CEO Josh James’s presentation at SES Chicago; the notion of the search marketer being the quarterback. Do you have any comment on why it is that the search marketer should be the one driving this process rather than someone else in marketing?
(This screen shot from Josh James’ presentation as SES Chicago reprinted with permission)
Chris Zaharias: Search has the largest, most measurable aggregation of buying intent that has ever existed. Because the opportunities for optimization are so much greater, the controls are at the fingertips of the advertiser. This is all important because you don’t get concentrations, of buying intent, expression of buying intent, and variations of buying intent in any other channel in the way that you do with search. So that aggregation of intent, the ability to analyze the data, the ability to optimize along multiple parameters, , that’s likely to be the path that the rest of marketing will take.
If everyone today is worried about Google becoming a monopoly, those worries are overblown because Google actually already is a monopoly for all intents and purposes within the world of search. It has an 80-95% market share in Europe. If you count their distribution partners, they’re almost at 80% in the States, and they keep taking more market share.
What’s really important to know is that the real time action marketplace that’s measurable and actionable within search is likely to grow into all other forms of marketing. We are a couple key technological advances away from advertisers being able to understand individual reactions to things like radio advertising, or print advertising, or billboard advertising. At which point, those offline marketing channels become measurable and actionable marketplaces where advertisers can react at the data they are seeing.
It’s much more likely that the search engines with the most efficient advertising platforms will grow into those other areas. And you are already seeing that with Google’s acquisition of DoubleClick, with some of the interesting tests they are doing around deployment and measurement of radio advertising.
When you see things like quick response codes in Japan, which led to tens of millions of consumers reacting to print advertising by scanning a barcode with their cell phone, it becomes very real to think about the systems in the action environment that is search. The role of the search marketer as a quarterback is not just important because it’s this biggest aggregation of buying intent right now, but also because the measurement, the real time optimization and the auction environment within search is likely to be the environment that’s subsumes the rest of advertising, not just online advertising.
Eric Enge: So even though the rest of advertising dwarfs search from an economic standpoint, the accountability and measurability of search will eventually become the predominant way of doing everything?
Chris Zaharias: Right, exactly.
When you think about the opportunities for advertisers to take advantage of the inventory that Google, Yahoo and Microsoft are aggregating, it’s pretty astounding. It is obviously incredibly complex, and it becomes all the more complex for every additional channel that you have to look at.
Eric Enge: Thank you Chris!
Chris Zaharias: Yes, thanks Eric!