Beyond Keywords: Multi-Variable PPC Targeting
Sometimes my colleagues ask me why I’m always requesting demos of new marketing technology. The answer is that I find major creative inspiration from the tools. As vendors compete in the marketplace, they’re forced to innovate to differentiate themselves from their competitors. For example, a few years ago, Yahoo developed “assists” in their search marketing […]
Sometimes my colleagues ask me why I’m always requesting demos of new marketing technology. The answer is that I find major creative inspiration from the tools. As vendors compete in the marketplace, they’re forced to innovate to differentiate themselves from their competitors. For example, a few years ago, Yahoo developed “assists” in their search marketing platform. This reporting provides users insights into all of the keywords are in a searcher’s path to conversion even if they’re not the final converting term. I’m sure when they were developing this in-house, it was an all around good idea, but I’d guess it was the fact they would be the first-to-market that was fueling the team. As a search marketer, being exposed to this tool literally sent my SEM point-of-view into new directions and now I always think beyond the last ad clicked—mainly due to Yahoo’s assist column in my reports.
Thanks again, Yahoo.
Recently, I’ve had another major shift in my SEM thought process due to another tool. I always considered the keyword as being the most granular level of my paid search programs. It’s always the pivot point in reports and certainly the engines seem to treat it as the “final level.” However, after being exposed to SearchRev’s bid management tool, I see now that the keyword is not basic building block of paid search. It’s the combination of the keyword plus targeting factors that should be the starting point of optimization.
Here’s a formula you can use to gain much more insight into how you should optimize:
keyword x engine x match type x geotargeting x dayparting x creative x landing page
Let’s take the term “online dating” as an exampe. The variation of “online dating” x Google x exact match x New York x Monday x the “10% off membership” creative x Promo Page #2 is a lot different than “online dating” x Yahoo x broad match x US and Canada x Weekends x “official site” creative x home page. If you just look at “online dating” for optimization, you’re really not looking at what’s really going on with that term. I would think that most SEM pros consider “online dating” exact match is different than “online dating” broad match and that each engine performs differently, but what about all of the other factors? Certainly one of those variations is going to have a better conversion rate than the others, require different bidding strategies, etc. The engines allow for this level of targeting, so why not use the technology to it’s full advantage?
SearchRev calls these unique keyword plus factor combinations “Precision Profiles.” Their bidding technology bids at the Precision Path level (called multi-variable targeting), not the keyword level. It’s this kind of thinking that has really begun to change the way I look at paid search—a medium I’ve been working with for years. I’ve always taken into account the various targeting factors in my paid search strategy. I’ve done dayparting testing, geotargeting testing, match type testing, creative testing, etc. I found the best converting factors and optimzed my accounts accordingly. However, it was always still at the keyword level. For me, that’s changed and I’m now onboard with Precision Profiles as the level of granularity I should be assessing.
Of course, creating campaigns with every geotargeting or dayparting variation in every engine is time consuming. Even though you know your bid of $1 for a keyword in New York gets you in 8th place but in Oklahoma City it puts you in 2nd place, the thought of generating fifty new campaigns for each state is painful to even think about. Multiply that by seven for each day of the week and you have three hundred and fifty new campaigns!
SearchRev’s platform is intriguing because it allows you to create multiple “shadow campaigns” very quickly. What this means is that inside the tool, you can choose various targeting options and the system automatically generates multiple campaigns in your different search engine advertising accounts. Even if you were to change the dayparting for one keyword, the tool creates an entirely new campaign in the engine for that variation. The end result is a very messy looking engine account, but the SearchRev UI remains intuitive. You can even split a combined campaign into separate content and search campaigns with one click.
I asked SearchRev’s CMO, Eduardo Llach, how this methodology developed. “This was an organic process.” says Llach. “We had initially created an infrastructure for complex search campaigns. Then, as we started testing things such as aggressive creatives (such as “free shipping”) versus conservative creatives (“$39.95 shipping”), we realized that different pools of traffic had different click-thru rates, conversion rates, etc. I have a background in mechanical engineering so we took an assembly approach of parent-child constraints to manage which combinations of factors worked best.”
The downside is that so many variations are hard to manage. You just can’t test everything. Imagine: 1 keyword x 3 match types x 15 geotargeting areas x 3 dayparting zones x 10 creative x 4 landing page is a staggering 5400 variations of one keyword. If you have even one thousand keywords in your accounts and are advertising on three engines, that’s 16,200,000 rows of data! Most multivariate testers will tell you that the rule of thumb is 100 conversions for each variation must be attained before you can achieve statistical significance and be confident in how well a variation is performing. It would even take Amazon.com awhile to get to that benchmark. So the solution is to not test everything but rather find interesting trends to test that you believe may have value.
SearchRev addresses this problem with automation to identify test segments. Llach says, “Our system actually looks at the very orders that come in and then builds campaigns based on those factors. We can also set thresholds for when to build these combinations by number of clicks, number of conversions, etc. So, for example, if ten orders all come in Monday morning in Boston on a keyword from Google, our system recognizes that and builds a campaign geotargeted to Boston, dayparted to Monday morning, and in the Google engine. That combination then gets tested and managed along with all of the other previous campaigns.”
I didn’t mean to make this post an advertorial for SearchRev, but their tool truly changed how I look at search. However, I think any search engine marketing professional will agree that the difference between success and failure in a highly competitive space like paid search is to find those little incremental percentage lifts. Going beyond keyword level granularity, multi-variable targeting, utlizing and bidding at the Precision Profile level are certainly some of my top best practices for finding those positive lifts.
Opinions expressed in this article are those of the guest author and not necessarily Search Engine Land. Staff authors are listed here.