Getting Core Paid Search Analytics Right (Part 1)
If you’re a diehard search marketer, you probably find web analytics conferences a bit strange. Many of the speakers are versed in the nuances of customizing third-party analytics platforms like Google Analytics, Omniture, and ClickTracks. But many of them seemed to have learned to sprint before they learned how to walk. I’ve been struck by […]
If you’re a diehard search marketer, you probably find web analytics conferences a bit strange. Many of the speakers are versed in the nuances of customizing third-party analytics platforms like Google Analytics, Omniture, and ClickTracks. But many of them seemed to have learned to sprint before they learned how to walk. I’ve been struck by how often search marketers have been told to pay close attention to some arcane point about an approach to data interpretation by people who have rarely, if ever, lived and died by the performance of an actual client paid search campaign.
While the “practitioner’s approach” may seem a bit more instinctive and less learned, the many detailed campaign improvements and tests that are the staple of day-to-day full time paid search marketing require the “instinctive practitioner” to be intimately familiar with the core—as opposed to exotic—workings of analytics and stats offerings. Thus they’re far less likely than “the Quants” to have gaping holes in their practical knowledge of the most obvious areas for improvement that need to be attacked on a regular basis.
For the sake of entertainment, let’s call the practitioners The Muddy Ones, after their tendency to get down and dirty. The legion of analytics experts with PhD’s or similar credentials: let’s simply call them the Quants. They’re both important to the profession, and Muddy Ones who fail to pay adequate heed to the power of formal math are bound to be overmatched in their professional challenges eventually. That said, it’s surprising how often the Quants are wrong about stuff, and how infrequently they focus on “quick wins” and “core tasks” of the sort that delight most clients and bosses.
Build a powerful campaign
The hallmark of performance-driven marketing is to manage campaigns to key metrics. But management isn’t marketing. So the first background point to keep in mind is that it’s all too easy to go through performance numbers and suggest responses. But how is that campaign built in the first place? What creative and planning processes go into it? Does the person building an AdWords campaign, for example, understand the way to architect a campaign correctly, in stages, so that management of that campaign becomes a meaningful exercise? There’s nothing worse than trying to “tighten up” the performance of an uninspiring, weakly-built campaign. No matter what you do, it’ll limp along at low volume at best.
A powerful campaign, as I define it, is characterized by (just to name a few):
- An understanding of Quality Score fundamentals: CTR and relevancy
- A model of how to test ads in different situations, and a plan for injecting response-inducing creative into the testing process
- A good understanding of filtering options and campaign settings
- A superior understanding of keyword research and keyword matching options
- Understanding ad distribution and reach
Managing everything to ROI
The convenience of tracking software like Google AdWords Conversion Tracker, Google Analytics, and (less convenient but also powerful) third party platforms like ClickTracks and Omniture means that there’s a consensus among the Muddy Ones and Ministry of Math alike that performance based marketing means reacting to conversion performance (or other KPI performance) and adjusting campaigns accordingly.
So that means you can adjust campaigns to keep any aspect or segment—a campaign, ad group, ad, individual keyword, segment of the content network, a time of day, and so on—within your target ROI range.
It doesn’t matter whether you’re a Muddy One or a Quant, this is where the debate actually gets rich in many cases. By ROI, what do we mean?
Muddy Ones know that rapid iteration is easiest within AdWords and similar platforms if you look at a target CPA (cost per acquisition) figure such as cost-per-order, cost-per-lead, cost-per-registration, etc.
But most of us would agree that on some other (perhaps looser) time frame, we need to look at ROAS (return on ad spend), involving real revenue or arbitrarily assigned revenue in the equation. That can lead to interesting questions related to data loss, time frames, repeat business, and lifetime value. Certainly, the initial customer acquisition should be given revenue credit in long-term planning. So by looking at longer-term data on true revenue associated with keywords and other segments, we gain added insight. The much shorter cookie life of Google AdWords Conversion Tracker means we can’t rely on it alone.
Assists are influential, but to what extent?
And of course attribution takes on even more “art-like” overtones when we start measuring assists. More and more, we’ll be asked to discount the weight of last-click keyword searches (such as a navigational seeming search for the brand’s website) and to give proper attribution to the banner impressions and research keywords that precede it in the buying cycle.
It’s easy to generate smooth talk about that type of issue, but for Muddy Ones, much harder to actually act on the information—especially when we rarely have full assist information today. Just because someone was exposed to some aspect of your multi-channel, many-keyword campaigns doesn’t mean that aspect contributed much to the sale. Does a keyword assist get 15% of the credit? 30%? One-off studies using control groups may have promising information (Quants love these because they are “walk-away” proof of concept), but to put it bluntly, they have little to do with the patterns of causation in the campaign you’re working on now. That’s someone else’s data, and you can’t verify either the consumer behavior or the related company financials. Nicholas Nassim Taleb would say that overreliance on other people’s studies is another example of The Luddic Fallacy (nerd fallacy, loosely translated).
Muddy Ones seem pretty instinctive in their behavior, but necessarily so. Because it’s so hard to see what’s “really” going on with every touchpoint, I’ve always felt that beyond the core CPA and ROAS numbers that we work with as a rough guide, attitude is everything. If you’re pro-advertising and bullish on the impact of your brand and website, then more exposure is good as long as it’s decently targeted. I think companies should be afraid of just stopping their advertising on their “on the fence, maybe helpful” keywords, just as they’ve always been afraid of the unseen effects of shutting down broad-based ad campaigns.
CPA tunnel vision: avoid
For rapid campaign iteration there’s no question that you should focus on something like a CPA number broken down by the appropriate segment, such as ad group or ad. Assuming you’ve created something that is worthwhile in the first place, iteration (bid management, ad selection) helps you fix what’s broken and choose the best paths.
But there are three problems related to this.
First, a powerful campaign requires experimentation to find right answers to questions. That includes ad testing and keyword experimentation. Overcautious management to CPA targets will lower volume and may thwart experimentation.
Second: more than lowering volume, there is a problem if your cautious CPA targets create much lower CTR’s than your competitors are generating. Plain and simple, the AdWords ranking algorithm loves high CTR’s. All else being equal, we need to try to build up CTR’s in an account whether we like it or not.
Third: Google’s black-box Conversion Optimizer product is one that automates the process of reaching target CPA’s. But it hampers hands-on tuning and learning on individual keywords. Advanced users should consider using Conversion Optimizer for set-and-forget functionality only once they’ve completed their own intensive process of gauging the performance of many individual keywords and matching options, to say nothing of ads. Third-party, alert-based bid-to-CPA tools are a wise use of automation as they allow the analyst to stay on top of wasteful situations while continuing to get down and dirty (Muddy) with hands-on optimization.
The most important point above was #2. Contrary to popular belief, CTR is a perfectly acceptable stat to include on your short list of core paid search metrics. Indeed, it’s vital, because it’s the most important component of Quality Score and will lead to higher rankings and improved volume for less money per click; thus, not only improved ROI, but more total profit.
Muddy Ones like to focus on campaign activities that tend to increase relevance and targeting. So when they use techniques to boost CTR on ads, segments of content, groups of keywords, etc., they use powerful techniques that ideally don’t harm ROI.
1. Improve matching options. Although Quality Score ostensibly normalizes for match type, I still feel that de-emphasizing broad matches and dialing up the emphasis on phrase matches is going to help your campaign on a number of levels. Not only will your CTR go up, but so should your conversion rate. Win-win.
2. Add negative keywords. Judicious use of negative keywords always improves both CTR and ROI. Always. Talk about a win-win. The only risk is damaging volume and profit by over-negativing. Don’t assume too much, just filter the obviously untargeted searchers and those with low buying intent.
3. Test ads religiously. We use a two-stage process that starts with “big wins” in terms of discovering behavior triggers for a particular customer base. At the refinement phase the idea is to use multivariate testing or other methods that will leave you open to discovering ads that create a “double win”—that is, ads that have significantly higher CTR, but harm ROI less than similar high-CTR ads (or in rare cases, actually improve ROI).
4. Find excuses to geotarget. Geotargeting is one of the hottest ways to target even more closely. You might even want to think of excuses to geotarget so you can raise account-wide Quality Score.
In my next posts on this subject, I’ll take a closer look at how to deal with imperfect conversion data, interesting segments to look at, and other core analytics functions that you can access without looking any farther than the Google AdWords interface. I’ll reinforce the above point about CTR’s vital role in successful campaigns. Finally, I’ll look at some fun ways to get more out of Google Analytics without having to pretzelize yourself into a Quant.
Opinions expressed in this article are those of the guest author and not necessarily Search Engine Land. Staff authors are listed here.
New on Search Engine Land