In part 1 of this series, I covered goals and the buying cycle. Today I'll talk about keyword opportunity and performance metrics to consider at various stages of the buying cycle and how you can derive actionable data from these.
What types of metrics should be considered at various stages of the buying cycle for each keyword? Essentially, all of the metrics you have access to, but my top ten follow (in non-linear order) along with the justification for Search Queries and Impressions as their role may not be so obvious.
# of Search Queries per keyword
It is important to understand the overall market volume for a search query (a keyword or keyword phrase) in order to understand trends and overall popularity. With this data you are able to compare market trends against your own performance for a term (either for an individual term, or a group of like terms) and then see if you trend over or under market. If you trend over market then this is a great sign and something that should be considered in optimization refinements. If you trend under market then you may want to consider changing your creative, upping your position/bid, and/or focusing your efforts on algorithmic results for that term or group of terms. Additionally, you may find terms at both the head and the tail of your relevant group of keywords that need to be included either because they generate so much pure visibility (simple eyeballs at the point of search) or for those at the end of the tail they may lead to the successful completion of a goal every time.
Data sources: adCenter Add-in for Excel (Monthly or Daily Traffic), Keyword Discovery, Word Tracker, Compete.
# of Ad Impressions per keyword
Similar to the importance of Search Queries the Ad Impression metric per keyword or group of like keywords enables you to compare market trends against your own performance for a term. It is a matter of preference, or depends on the goal at hand, which metric you chose to use – the Ad Impression number or the Search Query number. How are they different? Ad Impressions reflect the number of ads served per keyword or keyword phrase and reflect the match type variable that is relevant to you (broad, exact, or phrase) versus the Search Query data which reflects the number of searches performed for an exact term. The Search Query data is the pure reflection of the user's words entered into the search box and the Impression data is more of a reflection of the individual engine's keyword mapping choices. Both are important to consider when looking at how you trend to market and making decisions about optimization of your paid search campaign.
Data Sources: Engine reports for terms you are buying, adCenter Add-in For Excel (Monetization Data) for terms under consideration.
# of Clicks per keyword
Data Sources: Engine reports for terms you are buying, adCenter Add-in for Excel (Monetization Data) for terms under consideration, on-site analytics tools or ad serving tools
# of Clicks divided by # of Ad Impressions per keyword – Click Thru Rate
Data Sources: Engine reports for terms you are buying, adCenter Add-in for Excel (Monetization Data) for terms under consideration.
# of Visitors per keyword
Data Sources: On-Site Analytics Tools (Note: There are a number of free analytics tools out there in the market place, which have enough functionality for you to gather significant data if you're nervous about paying for a full solution. We have our own adCenter Analytics package which is in Beta at the moment. Learn more at advertising.microsoft.com.)
# of Unique Visitors per keyword
Data Sources: On-Site Analytics Tools
# of Page Views per keyword
Data Sources: On-Site Analytics Tools
Time on Site per keyword
Data Sources: On-Site Analytics Tools
Average Order Size per keyword
Data Sources: On-Site Analytics Tools
# and total $ amount of Latent Sales per keyword
Latent Sales represent purchases made by customers who have visited your site from a keyword search but did not purchase immediately, rather returned at a later session to complete the purchase.
Data Sources: On-Site Analytics Tools
# and total $ amount of Sales per keyword
Data Sources: On-Site Analytics Tools
What metric mix is likely to represent what part of the buying cycle? A very simplified suggestion for this follows, but each unique marketing scenario will lead to a new mix of meaningful metrics and buying cycle stages that need to be considered per keyword:
Research: High % of clicks without actions combined with a high number of page views, high number of unique visitors without single session actions, higher than usual time on site, # number of latent sales higher than other terms with direct sales
Consideration: High % of clicks with informational actions (newsletter requests, form fill-ins, etc…), combined with high number of return visitors, higher than usual time on site, higher than usual number of page views, number of latent sales higher than terms with direct sales
Acquisition: High % of clicks generating single session sales actions, high total sales amounts
Maintenance: High number of page views to informational pages, high number of maintenance related actions (could be owner newsletters, etc…)
How do you make sense of all this data? How do you find information from thousands, or hundreds of thousands, or millions of keywords and associated attributes? To do this I recommend that you classify your keywords according to a taxonomy that resonates within your organization. Ultimately, creating a custom taxonomy will help you share your findings meaningfully. Pivot on the metrics that you have available by the classification system you have created to see how groups of terms are performing. This will help drive actionable recommendations for each stage of the buying cycle that can be easily approved within your organization – because they will be presented in a language common to your audience.
When you begin to consider the value of keywords prior to the point of conversion it will likely change your bidding strategies, or your decision to keep bidding on a term, or not. This will provide a more complete picture of your keyword landscape and the importance of one keyword over another – plus if you can see the search funnel data you can begin to learn how keywords work together holistically across both paid and natural search in addition to the time continuum. One of the benefits of looking at your data in this manner is to pull out highly valuable terms from the keyword long tail. If you are optimizing based on number of sales, or total dollar amount sold per keyword then you may miss keywords that have a high conversion rate per click all the time, even if you only have one or two clicks on that term a month.
Implied in this has been that the notion of keyword analysis from both paid and natural search. You may want to look at one or the other, or compare the two against each other versus looking at the aggregate. The same concepts apply for your analysis whether you are looking at paid, natural, or search traffic as a whole.
In the end, we have to define goals to help us make any decisions for improving our work. If we recognize that our goals may be different for each stage of the buying cycle, or for different types of campaigns then we have an opportunity to improve the efficiency of our efforts. It comes down to defining the goals, understanding the available metrics, reviewing the data iteratively for each goal set in its own terms, and ultimately understanding how optimizing for different goals helps improve what is the uber-goal: Sales (in most cases).