Posted by Eric
August 23, 2013
Quadrant Analysis of Google AdWords Performance
The idea is incredibly simple… get your product or service in front of people searching the internet for relevant terms or concepts. Unfortunately, the simplest ideas gain complexity as they grow and change. The same is true for search engine marketing, pay-per-click advertising, etc. This blog will focus on Google AdWords, but the approach is valid for any vendor.
The entire purpose of analyzing AdWords is to answer the basic question… “How is my expenditure on AdWords performing?” or “is my AdWords investment worth it?” As basic as these questions are, the process of answering them is anything but simple.
The first problem… which of us has used the AdWords interface and seen this incredibly overwhelming magnitude of data?
That would indicate over 14,000 entries to be evaluated for performance! Enter Unilytics’ AdWords Performance Quadrant… a quick and easy way to identify good, bad, or neutral performance in keywords with very little effort.
A good start in dealing with all this data is to extract the data and chart it, for example, using Tableau Desktop, we could create the following:
In this example, we show conversions along the horizontal axis and cost along the vertical. These are the most important measures as “conversions” is what we’re trying to generate, and “cost” is what it takes to get there. Unilytics’ AdWords Performance Quadrant uses these two basic measures and lays them out in quadrants on a chart.
When looking at the AdWords Performance Quadrant, each quadrant provides highly intuitive, rapidly interpreted information. In our example, the quadrant describes the AdWords campaigns as follows:
In this example, the following interpretations can be made:
It’s not only important to determine which quadrant an item is in, but how its position in that quadrant has changed over time. Using the “paging” feature of Tableau we can determine the velocity of change over time for these campaigns. For example:
This shows that the campaigns all started off similarly, and performed similarly for a period of time at the beginning, but then rapidly spread apart from each other. This chart has been manufactured to exaggerate the point we’re making in this blog, but the theory works even in a real-world situation.
Speaking of which, let’s look at some real world data using the AdWords Performance Quadrant:
In this case there are no “bad” campaigns running (i.e., upper left quadrant); one expensive, but productive campaign (i.e., the “Toronto Crime Academy” campaign in the upper right quadrant); two “great” campaigns running in the lower right quadrant. All the rest of the campaigns may be too early to tell at this point.
Let’s now take a look at the “rate of change” for those two “great” campaigns:
For both of those campaigns, the performance is pretty consistent over time. In other words, there’s no extreme, unexpected change in conversion rate or cost.
Let’s look at another example, and answer two questions at once. Question 1, when can we make decisions about items in the “too soon” quadrant on the lower left, and question 2, what does an “extreme” change look like using this approach? To answer both questions, look at the following chart from the real world data:
In this case, the “Top Winnipeg Criminology School” started off fairly normally, like most campaigns, but the cost suddenly started to escalate without a related increase in conversions. Perhaps a competitor jumped on certain keywords and increased their maximum bid? While we don’t know the “why” the increase happened, we can see that it happened.
In this example, this is a bad sign, and possibly makes the campaign a candidate for defunding, reducing the maximum bid, or changing the strategy of the campaign to return to a normal growth trend.
Keep in mind, this approach works well for any level of the AdWords hierarchy, including networks, campaigns, ad groups, ads, keywords, etc. These examples were based on campaigns, but are equally useful for the other levels of AdWords information.
To summarize, the goal of this blog is to give you tools to help reduce the headache of managing 14,000 keywords, campaigns, ads, etc. We’ve shown a simplified version of the theory of the AdWords Performance Quadrant, as well as a real-world example. Give this approach a try and we’re sure it will help you differentiate “the signal from the noise” and make you an AdWords rock star!
If you need assistance obtaining your AdWords data, configuring your charts, or evaluating and analyzing your results, feel free to contact us. Unilytics highly recommends using Tableau Desktop for implementing the AdWords Performance Quadrant due to its ease of use. Please take a look around on our site for additional information on Tableau Desktop and why Unilytics so strongly recommends it for this type of work. You can download a Tableau Desktop brochure here.
Also, keep in mind that Unilytics has extensive experience crunching other AdWords data and providing entire frameworks of intelligence and decision-making capabilities. This blog only covers one of numerous dashboards useful to those administering an AdWords program. For example, other dashboards we use and configure for clients include:
If you’d like to see examples of these other AdWords performance dashboards, feel free to drop us an e-mail at firstname.lastname@example.org and we’ll get them to you and even arrange a demonstration if you like.