Search, whether SEO, paid search, or site search, is a word problem.
While in the previous chapters we have looked at some mathematical formulas you can use to capture and track the results of your various search activities, search is really a very complex word problem. When you think about it, it’s actually pretty amazing what the search engines do, how well they do it, and how frequently. Consider how much content the search engines have indexed. There are over one trillion unique URLs on the Internet (http://googleblog.blogspot.com/2008/07/we-knew-web-was-big.html), and from that content, the search engines are able to pluck out the best page that represents a given term in the blink of an eye. When I put the term “coffee” into Google’s search box, it returns that result in 0.16 seconds—that’s pretty impressive (Figure 4-1). Not only that, but the search engine also maps where the nearest coffee shops to me are located. Unreal.
Suppose your website is all about coffee. If one of your pages appears in the results list when someone types “coffee” into a search engine’s query box, you have already accomplished a lot. As you can see in Figure 4-1, there are some 230,000,000 other pages also competing on this one word. That’s a lot of pages about coffee! When you see those results you may say to yourself, “But I’ve got the best page ever on coffee, and Wikipedia ranks number 1?” Yes, yes it does, and for a number of reasons (with the biggest most likely being the number of inbound links that page has and the authority of the site).
For organic rankings, the search engines use both on-page and off-page indicators. If you feel you should be present on this word, I suggest now is the time to begin a paid search campaign. As you set up that campaign, ask yourself, “How much is this click worth to me?” In this chapter, we’ll be looking at tracking the words you think you should be ranking on so you can decide if you have a problem (say, if Wikipedia is ranking ahead of you), or if you are lucky and already rank very well on those terms. We will also be looking at how to tell how much you are paying per word, and if you feel it is worth more or less.
Words and terms people use to find your content that result in positive ROI and engagement are ultimately what you want to be measuring. There are a number of ways to track words, and the strategies for SEO and paid search vary in terms of how you will want to manage your word lists. In this chapter, we will look at determining which are your top-performing words and which are underperforming, deciding what words have potential (do you really think you will rank on “coffee” if you sell tea?), clustering words around your pages, and creating whitelists and negative keyword lists. We’ll also explore some specific metrics to look at for paid search campaigns.
Lastly, when selecting your words, you will need to use some judgment as to whether a term is really relevant. Just because you have the choice of tea or coffee in the morning doesn’t mean pages about tea should rank in a search for coffee, and vice versa. When people search for a term, they want to find the pages that are most relevant to that term. The search engines rewards you for relevancy: if you get to the top of the rankings and the engines start to notice a lot of people coming back and searching on the same term again, the odds are they will pick up a signal that your page is simply not relevant. Do you really feel like spending all that time to get to the top of a search only to find you have a high bounce rate? I guess that depends on whether you can make a profit that way, but generally speaking it’s not ideal.
Tools you will need in this chapter:
Clickstream tracking package (Google Analytics, Adobe SiteCatalyst, etc.)
Google Keyword Tool or some other keyword research tool
Spreadsheet program (Excel or something similar)
AdWords or some other paid search tool
Keyword tracking program (e.g., WebPosition)
Understanding what words are driving people to your website—and more specifically, what words are driving people to a particular page on your site—is the first step to understanding opportunity. There are two components to consider: what words currently drive traffic, and what words you think should drive traffic. Further, the choice of words you use on a web page can impact SEO factors on that page.
As we saw in Chapter 2, when looking at ROI, improving the ranking of a page on one term can impact the volume of traffic to the site, which should drive incremental ROI. Typically, when optimizing a site for SEO, there are site-wide tactics and page-level tactics. Page-level tactics usually involve mapping keywords to a page and then saying that page should rank on those words.
On more than one occasion, I have been asked how to improve the ranking of a page on a very aggressive word. Particularly when dealing with large enterprise sites, you can end up with several departments and teams all competing over the same keywords, each with its own page.
This quickly becomes counterproductive, as instead of working together to build a unified front, these organization are fracturing and splitting their SEO efforts to see minimal gains. At an enterprise-level, if you plan on targeting an aggressive keyword, you cannot expect the search engines to reward you with relevant rankings if you can’t send positive on-site signals (such as well-developed inbound links and site information architecture), let alone trying to build links off-site.
If you can’t fix the things you have direct control over, you can’t ever expect to rank on a high-value nonbranded term. Experience has shown me that sometimes it’s not a technical issue that needs to be resolved but an organizational thinking and strategy issue, especially when bonuses are based on success and the bonuses force organizations to fragment their SEO strategies.
This approach of assigning words to a page seems a bit backwards to me, in that you should know what top-level words are already driving traffic to a page. You can find this out by examining your clickstream data and building a keyword cloud or keyword cluster.
To develop our keyword clusters, we must first identify a page in our clickstream tool. As a heads up, when you run this report, you may find that the words that drive traffic to your page are not the words you think should be driving traffic to that page. The top words should be relevant to the content.
In Google Analytics, our first step would be to find the page we want to optimize under Content→“Top Content” and click on it. Under “Analyze,” select “Entrance Keywords” and you will be presented with a list of all the words that have driven traffic to your page (Figure 4-2). To filter only for the organic words, click “Advanced Segments” in the upper-right corner and select “Non-paid Traffic.” In Google Analytics v5, you can generate a keyword cloud by changing how reports are visualized when viewing the keyword report—you can also filter based on page, if you operate on a more visual level.
You should be able to see the number of page views, number of unique page views, bounce rate, exit rate, and average value of the page. Table 4-1 shows a sample of the data from Figure 4-2. Also take note that there are over 130 terms driving traffic to this page. Remember when we talked about the long tail? This is a perfect example of this in action. The top term, “playday.com,” drives only 29% of the traffic; all the other terms combine to create more traffic than that one term. Optimizing that term in ways that impact the other terms may have dire repercussions. Also surprising is the volume of searches on “skydiving in Brandon Manitoba”—a term that most likely was not intentionally optimized for, but a term that drives some significant traffic on its own.
Table 4-1. Adding in page views to keyword clustering
|Keyword||Page views||Unique page views||Bounce rate||% Exit||$ Index|
|activities in Canada||72||33||0%||43.06%||$0.00|
|skydiving in Brandon Manitoba||36||18||0%||50.00%||$0.00|
|recreation in Canada||26||12||0%||50.00%||$0.00|
Looking at this data set, we see that we don’t have an indexed value to derive an ROI value from. The real world will often throw us curveballs like this. For this case, I would suggest estimating an ROI value by determining which words achieve success based on other measurable activities. For example, which words deliver more traffic to a page with a high-value goal or activity? This would be an ideal time to use a pathing report and filter by individual keywords. This process can become laborious, and you may choose to look only at the top 10 words driving traffic to your site, as they will likely be the highest sources of revenue. You’ll want to build your strategy to revolve around the words with the highest conversions and goal completions, relative to the customer’s stage in the conversion funnel.
Once we know what keywords are driving traffic to our home page, we can also better decide if perhaps we should look at pushing some of this traffic into other sections of our site, by modifying links on the home page, utilizing some of these top keywords and pointing further into the site. By doing this, we indicate to the search engines that our site contains another page on this topic. If we build up enough of these links on our site, we may be able to transfer that traffic from the home page to a deeper page that may be better able to serve the needs of the customers, improving both traffic and bounce rate. We only want to do this if we expect to see a positive ROI on this effort, though. We can take this data and use it to project which words would be the best to try to further optimize and move up the rankings, to bring in more traffic.
Through our keyword clustering, we can also look at segmenting words. There may be groups of words that can be further grouped together: for example, in Table 4-1, we see that we could potentially segment our keywords based on geography or activity.
The concept of clustering keywords around pages was first introduced to me by Bill Hunt. He’s a super-smart guy, and has really led the way in the paid search and SEO field. What I like about this concept is that it makes you realize that you are not looking at a one-page–to–one-keyword mapping; instead, you are looking at sets of words that play off each other.
If you think about mapping one page to one keyword, you end up very focused on a short-head objective. By building up keywords in clusters, you start to think of words in relation to each other, and you start to think about the midtail and beyond. Further, it is much more manageable to think of keywords in terms of concepts and ideas and relevancy. My spin on this concept is that I like to think of one lead word that is the target word, and several supporting words around that word. You want to maximize your optimization to cover all those words, but you may find the cluster has more value for increasing ROI than the single lead word does.
The lead word keeps you focused on the overall concept and idea of the page and helps you quickly sum up what the page should be about.
Segmentation helps to make large groups of words more manageable (e.g., as in Figure 4-2, where we have 134 different terms driving people to one page). By segmenting words in such a way, we can refine our CRO strategy by looking at whether each of these segments has appropriate space on the page and can serve the needs of people likely coming in on these search results, and considering whether it would be appropriate to split the content across several pages rather than leaving it all on one page.
You also want to find out how these words currently rank, as well as the specific page in the search results. To find your ranking, you can use WebPosition (Figure 4-3) or any other rank checking program.
We can also look at segmenting our words into long tail and short tail terms; that is, our top 5 or 10 driving terms and then the subset of terms that drive further traffic. This can help us identify terms that are positionally weak in the rankings. Table 4-2 shows the SERP positions for each of the keywords we identified earlier.
Table 4-2. Keyword clustering against our homepage
|Keyword||SERP position||Page views||Unique page views||Bounce rate||% Exit||$ Index|
|activities in Canada||5||72||33||0%||43.06%||$0.00|
|skydiving in Brandon Manitoba||8||36||18||0%||50.00%||$0.00|
|recreation in Canada||7||26||12||0%||50.00%||$0.00|
By adding in the SERP position, we can identify the terms where we have the greatest opportunity to improve traffic. We may also find that some of these terms are very difficult to compete on.
Creating our keyword cluster has given us a better idea of what exactly is happening with this page in terms of its SEO value. We can see both the short head and long tail impact, and what terms convert well and do not convert well. Also, we now know which terms have the greatest room for improvement.
Knowing the potential of these terms will help us to build out our ROI formula to figure out which of these terms are the ones that will have the greatest impact on our ROI. The words that will generate the greatest revenue changes are the words that we should target first. In this case, because we do not have an average customer value, we will assume the goal of the site is simply to drive traffic to generate advertising impressions. In Table 4-3, we include another column called “Potential traffic increase.” We will use what we know about ranking position to estimate how much additional traffic we could drive to each section by moving each term up three spots or to a number 1 position (whichever is less). This doesn’t mean these changes will be easy to achieve. To measure feasibility, we will have to build out a competitor analysis, which we will cover in Chapter 8.
Table 4-3. Adding in growth potential for keywords
|Keyword||Avg SERP position||Page views||Unique page views||Bounce rate||% Exit||$ Index||Potential traffic increase|
|activities in Canada||5||72||33||0%||43.06%||$0.00||53|
|skydiving in Brandon Manitoba||8||36||18||0%||50.00%||$0.00||23|
|recreation in Canada||7||26||12||0%||50.00%||$0.00||20|
To figure out the change in traffic volumes you are likely to see if you move up or down a position, use the following formula:
Expected Total Page Views – Current Page Views × (1 + Projected SERP CTR % Converted to a Decimal)
Finally, to calculate the potential traffic increase, use:
Potential Traffic Increase = Expected Total Page Views – Current Page Views
Here, we can see that our best improvements in traffic volume will come from improving the rankings of the terms “playday” and “activities in Canada.” Knowing this, we can now start to explore how difficult it would be to improve these terms in the rankings based on an analysis of our competitors (we’ll return to this in Chapter 8).
Another analytics tool you might want to consider is Wordtracker Strategizer. It provides insights into the click-through share you have on a word, as well as an idea of some of the better opportunities you may have for improving your ROI. Strategizer imports Google Analytics keyword reports and also incorporates Google AdWords data to give you an idea of the click share you have per word (http://www.wordtracker.com/academy/automatic-seo-case-study-tahoe-mountain-sports). Strategizer allows you to look at the data in a number of ways, so you can figure out if it is worth going after terms with high volumes and low conversions or high conversions and low volumes. You may still need to do some of your own math, but Wordtracker is working on making the tool simpler to use and making it easier to spot opportunities. With the integration of Google Analytics, Wordtracker Strategizer has the potential to be a very powerful tool in your kit if you take the time to sit down and understand it. It may save you significant time trying to figure out what words have the best value.
Understanding what we want to optimize on for currently ranking words is a start in identifying opportunities, but we also need to look at what we are not ranking on, and if it is worth investing time and effort pursuing rankings on new terms or words. This requires keyword research.
No tool will tell you exactly how much traffic will come from a word. I have used the Google Keyword Tool for research on keywords and been told that the traffic volume is too low to measure, and yet in my clickstream data, I see significant volumes of traffic from these terms. This means the keyword research tools are good for estimates only.
The Google Keyword Tool was designed to support paid search activities, but it can be leveraged for organic research as well. The tool provides some insights into search volumes (Figure 4-4). When using the Keyword Tool for organic keyword research, make sure that you always have [Exact] match selected. This option will only be available (on the lefthand side) after you enter the words you are looking for. Selecting [Exact] will provide data for a specific search on a specific term.
The Keyword Tool also allows you to filter based on geography and language, as well as device type, should you want to see the difference between searches on desktop and mobile devices. The tool provides global and local monthly search statistics, so you can get an idea of a specific geography, as well as the global volume if your brand reaches across borders.
Using the tool is very simple: just put in the words you want to learn more about, and it will return information on those words, as well as suggesting other similar search queries. With the estimated traffic volume you can project a volume of traffic based on targeted position. Take the estimated volume and multiply it by the percentage based on organic ranking position to determine approximately how much traffic opportunity may exist for you on a word or term.
As an example, consider the word “cereal.” If it has roughly 34,000 searches a month, and you feel you can capture the 8th ranking spot, you are likely to receive 1190 visits from that word (34,000 × 0.035 = 1,190). Based on your site’s average conversion rate, you can then project approximately how many of these customers you expect to convert to a sale. If we assume a 3% conversion rate, this means 1190 × .03 = 35.7 conversions. Knowing you would have 35.7 conversions, you can project an approximate revenue by multiplying the number of conversions by the average order value.
It will likely take more effort and time to capitalize on opportunities for keywords that you do not rank on, which you will also have to take into account. Most of the time, you will find yourself looking to improve organic rankings for terms you currently rank on, which makes sense. As you already have an established traffic base, and some insights into behavior, you can compare how you currently rank to how your competitors rank and determine if you have a chance of beating those competitors. Further, you have the opportunity to optimize your page for traffic already coming in on that term, meaning that even with the same traffic volume, you can improve conversion rates and on-site activities as you work at pushing up the rankings.
Working to rank on terms you don’t currently rank on will be a much more involved process, and you may want to consider looking at paid search to bridge the gap and build awareness of content you have that is related to certain terms. To understand how to improve rankings on new terms, you need to understand how people find you today.
The first question you should be asking is how people are finding your site today. What words are already driving traffic to your site? What words might a search engine see as the most relevant to a page that you’re developing? These are important questions, and ones that can be answered. First, though, you should understand some basic theories around the search engine algorithms. The terms latent semantic indexing (LSI) and latent Dirichlet allocation (LDA) are often bandied about when looking at keyword relevancy in a document. LDA can be seen as a sort of extension of LSI. The goal of both is to provide some semantic context to terms, words, and groups of words. In layman’s terms, the search engine algorithms try to provide context to words, to understand synonyms, and to recognize that some word pairings have different meanings when together than they do separately.
The engines try to create an association or understanding of the context of the words based on use patterns they see. Without knowing for sure, SEO people need to guess at which models best match what is used by the engines. LSI has been ruled out, as it doesn’t scale well. LDA is a possible option, as is the Hidden Topic Markov Model (HTMM). Then there is certainly a great likelihood of something unique to each subsystem. In Google’s case, this may be its proprietary “phrase-based information retrieval” system. The larger body of work (the Internet) and all the data it contains also plays a big role in the algorithm.
In a Wired article published in 2010, Google admitted to using user behavior as a way to “educate” its algorithm even further about variations in synonyms; the article also discussed how important the proximity of words to each other is to understanding the differences in context of a set of words. So, what we do know is that frequency and context within a page play a role in determining on-page factors. For example, use of the term “hot dog” alongside terms such as “ketchup” and “relish” indicates a food-based context. The algorithms used in both Google and Bing are much more sophisticated than simply looking at how many times a word appears in a document. Feel free to go out and learn more about these topics if you are interested in understanding search patterns. For our purposes here, we do not need to get any deeper into the algorithms. The point is that, as we look at words and groupings, we may see some oddities in words linking to pages; also, when using keyword density tools and keyword volume tools, we should bear in mind that they are simply very basic tools that can provide some limited insight into how frequently a word appears in a page.
Knowing that there are algorithms that look at synonyms and other word variations, we do have a limited set of tools we can use to measure relevancy to a specific word. There are several options for measuring keyword density and volume, as reviewed in the Appendix A. Each of these has its pros and cons: the value of a keyword density or volume checker is not to check the relevancy of a term on a page, but to help identify if there is a word or term that is significantly over- or underused. When trying to determine the relevancy of a page using an LDA tool (it may be worth keeping an eye out for the one SEOmoz has in the works (http://www.seomoz.org/blog/content-optimization-revisiting-topic-modeling-lda-our-labs-tool)), you may use a density check to see if a certain term may be in use that is bringing down your LDA score. Using the LDA score to determine relevancy and then a volume or density tool to check the frequency of terms in a document can provide you with much better insight into the differences between your page and a page ranking higher than you in the search results on a specific term.
For our purposes, we want to know what words are bringing people to the content on our site. Secondly, we want to know what the likelihood is of any new content we create being found. That latter point is key, as it allows you to start thinking about the search process before content is created. If you can enter this mind-set when creating content, you will start to realize that it’s not all about keywords; including images or video that can also be indexed may help drive traffic and improve the findability of your site and content. You will also come to really understand how important the long tail is compared with the short head in terms of driving traffic by the end of this chapter.
For organic optimization you will also need to consider how you will build awareness of new pages to generate new links from other sites, as well as raising social awareness and building social links to your site. These are all off-site factors that significantly help improve rankings, and they are all things we can measure. We will look at these issues further in Chapter 9.
For paid search, we want to know which words are driving traffic that converts, and how to maximize good terms and minimize bad terms. Essentially, we will be creating “whitelists” and “negative keyword lists”—terms that should be familiar to most paid search folks—and setting up proper tracking and auditing of these terms. We also want to be able to predict seasonality effects and understand when some words will go big, and we will look at building keyword clusters for our paid search pages, as we did for our SEO pages.
Paid search is a different beast from organic search when it comes to making decisions about keywords. When you’re setting up a paid search campaign, there are different types of matches you can set up for keywords, and these match types differ between Bing and Google. The following sections provide definitions of these match types directly from the engines and examples of when ads will run. Because there are different types of matches and they are interpreted in different ways, a given word can be either a drain on your budget or a great revenue generator, depending on the match type. Match types will impact how quickly you burn through money, and how targeted your ads are.
Google approaches broad match as follows:
This is the default option. If your ad group contained the keyword tennis shoes, your ad would be eligible to appear when a user’s search term contained either or both words (tennis and shoes) in any order, possibly along with other terms. Your ads could also show for singular/plural forms, synonyms, and other relevant variations.
Bing, on the other hand, handles broad match a bit differently:
[It] triggers the display of your ad when individual words in your keyword appear, in any order, in a customer’s search query. For example, your keyword red flower would match search queries that include red flower, flower is red, and other variations, and not just red or flower.
Broad match can expand to include words that are closely related to your keywords. For example, a search query for red carnation might result in your ad being displayed, because adCenter automatically identifies carnation as a type of flower. Use broad match to expose your ads to a wider audience.
Table 4-4 compares the way broad match is handled by Google and Bing.
Phrase match is handled by Google in the following way:
In this area, Bing is quite similar to Google, as illustrated in its documentation:
[It] triggers the display of your ad if the word or words in your keyword appear in a customer’s search query—even if other words are present in the typed query. Your keyword red flower would match searches for big red flower and red flower, but not yellow flower or flower red.
Table 4-5 shows a comparison of how phrase match is handled by Google and Bing.
Google explains its approach to exact match as follows:
Bing’s approach to exact match is somewhat similar, though it does differ in some subtle ways:
[It] triggers the display of your ad only when the exact word or words in your keyword, in exactly the same order, appear in a customer’s query. Your keyword red flower would only match searches for red flower, with no spelling variations. With exact match you might see fewer impressions but a higher click-through rate, because your ad is shown to people who might be more interested in your product.
Further, Bing seems to ignore words like “the”, “a”, “an”, etc. Table 4-6 provides a comparison of how exact match is handled by Google and Bing.
Google allows for different match types on negative keywords (http://adwords.google.com/support/aw/bin/answer.py?hl=en&answer=67991):
- I. Negative Keywords
You can add negative keywords at both the ad group level and the campaign level. Adding a negative keyword at the ad group level means that the term will only affect the ads in the ad group. A campaign-level negative keyword will apply to all ads in all ad groups in that campaign. [...]
For example, adding free trial as a negative keyword to your account would prevent your ads from showing on any search queries containing the terms free and trial. It wouldn’t prevent your ads from showing on variations of these terms, however. It also wouldn’t prevent your ads from showing on search queries that only contain one of the terms.
For example, the search queries one-day trial and free test could trigger your ads, while free one-day trial could not.
- II. Negative Phrase-Matched Keywords
You can create a negative phrase-matched keyword by surrounding the term with quotation marks. Here’s an example:
If you were to add “free trial” as a negative keyword to your account, the system wouldn’t let any search query containing the phrase free trial trigger your ads. The search query free trial lesson would not trigger your ads, for instance. The rules of phrase match still apply, however, so your ads could possibly show on the search query trial or free one-day trial.
- III. Negative Exact-Matched Keywords
You can create a exact-matched keyword by surrounding the term with brackets. For example:
Adding this as a negative keyword would prevent your ads from showing on the search query free trial only. Search queries such as free trials, free, and one-day free trial could still trigger your ads.
It’s a good idea to add relevant variations of your negative keywords, including both the singular and plural forms. [...]
Note: When adding keywords directly in the negative keywords section, there’s no need to include a negative sign (-) before each keyword.
Bing only has one match type for negative keywords, which is phrase match (http://community.microsoftadvertising.com/blogs/advertiser/archive/2010/10/25/feature-comparison-series-match-types-and-negative-keywords.aspx). However, you can apply negative keywords at different levels, with different results depending on the level, as outlined in Bing’s documentation:
Campaigns. Campaign-level keywords apply to all keywords in a campaign, unless you also associate negative keywords with an ad group or a specific keyword. Each campaign can contain thousands of negative keywords.
Ad groups. If you associate negative keywords with an ad group, campaign-level negative keywords will not be applied to that ad group. Each ad group can contain thousands of negative keywords.
Specific keywords. If you associate negative keywords with a specific keyword, campaign-level and ad group-level negative keywords will not be applied to that keyword. Each keyword can have a list of negative keywords of up to 1024 characters, including commas.
Table 4-7 compares how “black kittens” is handled as a negative keyword in the two search engines.
Table 4-7. Black Kittens as a negative keyword
|Search query||Google negative broad match Black Kittens||Google negative phrase match “Black Kittens”||Google negative exact match [Black Kittens]||Bing negative match Black Kittens|
|Free Black Kittens||no||no||yes||no|
|Kittens that are Black||no||yes||yes||yes|
Broad matching in paid search is a very useful way to build up new word pairings to help improve your impressions across a variety of words. Broad match is defined by Google (http://adwords.google.com/support/aw/bin/answer.py?hl=en&answer=6136) as follows:
With broad match, the Google AdWords system automatically runs your ads on relevant variations of your keywords, even if these terms aren’t in your keyword lists. Keyword variations can include synonyms, singular/plural forms, relevant variants of your keywords, and phrases containing your keywords.
For example, if you’re currently running ads on the broad-matched keyword web hosting, your ads may show for the search queries web hosting company or webhost. The keyword variations that are allowed to trigger your ads will change over time, as the AdWords system continually monitors your keyword quality and performance factors. Your ads will only continue showing on the highest-performing and most relevant keyword variations.
When you bid on a set of words on broad match, you can also generate a report of what words your ads ran on. To get this report, simply follow these steps:
Sign in to your AdWords account.
Click the Campaigns tab.
Click the Keywords tab.
Check the boxes of the specific terms you want to see that drove traffic, or leave all unchecked to see all words that drove traffic.
Click the “See search queries” button above your statistics table.
Select “All” from the menu to analyze the search queries for all your listed keywords.
Running this report will help you understand what actual words are driving the clicks through your AdWord campaigns. You can use this information to create lists of positive and negative keywords. Say, for example, you’re a cat breeder and you’re selling cats. Suppose the current litter of kittens is orange. When you’re starting out, you may run a broad match campaign on “cats” to learn about what words people are using to find cats. Suppose you discover that “grey cats” and “tan cats” are driving most of the traffic. Since you don’t have any grey cats, you’ll want to put “grey” on your negative keyword list. This will guarantee that any searches with “cats” and “grey” in them will not display your ad.
To this list, you’ll add all the terms you do not want to show up on, effectively saving you money by ensuring you don’t rank on nonrelevant terms. “Tan cats” is a relevant description for your “orange cats,” so you may want to capture people looking for “tan cats.” You don’t have to do anything to make sure your ad shows up on these searches, but you may want to consider creating a new ad specific to this term to run on exact or phrase match. By creating a more relevant ad with relevant copy, perhaps better describing your tan cats, you may be able to capitalize on conversions.
The negative keyword list will be important for filtering out words you do not want to spend money on, but you will want to continue to run the original keyword on broad match if you want to keep learning other terms that are used to find cats—essentially, data mining the results that trigger clicks to your site. We can see an example of setting up a negative list in Figure 4-5. We can apply negative keywords both at the campaign level and the ad group level in a variety of ways that we will not get into here, as we are focusing on analytics and not list setup.
Based on the data you get from your search queries list, you should be able to quickly build up a negative list of words that you don’t want to show up on. The sooner you build up this list, the sooner you can stop spending money on clicks that have little to no value to you.
The other source of negative keywords is your clickstream data. If you are tracking the volumes of traffic and conversion rates from your search terms (and if you aren’t, what are you doing?), you should be able to use the clickstream data to identify words that are not performing in terms of creating conversions or that have poor or negative ROI.
These are words you may want to add to your negative list. By weeding out underperforming words and irrelevant terms, you will increase the likelihood that your remaining keywords will rank well on broad matches, driving traffic to your site and creating positive ROI for you.
You will also need to make some decisions about the words you keep in your keyword list. For example, you will want to consider what words should be set to [exact] match or “phrase” match so you can keep better track of these terms and how they perform. You can create more targeted ad groups specific to these sets of words, and once you have an understanding of which words you want to show up on for exact or phrase matches, you may even decide to turn off your broad match words.
If you decide to go from broad to exact match, the effect is essentially turning everything that you are not matching as an exact phrase into a negative list. Your keyword list will also likely have a higher CTR as you become more precise in your ad messaging and start to display more relevantly on more controlled sets of words, and you should start to see higher conversion rates.
Running on broad match, though, is a good way to identify words that have the potential to drive or influence your paid search campaigns in the long term. You may discover words you hadn’t thought of that turn into top-performing words, which is ultimately what you are striving to develop. Your goal should be to build up a set of top-performing words while removing any words that underperform by placing them on your negative list.
To track top-performing paid search words, much like we did on the SEO side, we will need to develop a dashboard, spreadsheet, or matrix that tracks our words along with some KPIs that are important to each of those words. We want to track our average visitor value, cost per acquisition, and ROI. All of this data is available either through your clickstream data or through your keyword management tool, if you’ve set it up to capture conversions. By comparing our cost per acquisition and our average customer value, we will be able to determine if we are running a negative or positive ROI. Table 4-8 shows the kind of table you will want to build up.
When we look at our ROI, we can see that some words are performing better than other words. You will likely discover that 80% of your revenue comes from 20% of your keywords. When looking at your keyword list, you should work down from highest ROI to lowest ROI, ensuring that your top-performing words have full coverage. This will help us plan our budget. If you do not have full coverage on one of your terms that has a high ROI, you may consider pausing a lower-ROI term to maximize exposure of the stronger ROI terms.
When a campaign first starts, purchases from early clicks will result in an extremely high ROI. You should wait until you have enough statistical data to make an informed decision before you start making changes—usually three to four weeks of data is enough, though if they are low-volume terms, you may need to wait longer.
Table 4-8. Paid search top-performing and underperforming words
|Keyword||Cost per click||Visits||Conversions||Average visitor value||Cost per acquisition||ROI|
The top-performing words should be rewarded with further investment, while the underperforming words should be paused and reviewed. There may be reasons why these words are not converting well. For example, the words may be more costly due to poor Quality Scores, or they may have poor bid strategies or high competition. This is not to say you should give up on these words; what you should do is investigate if there is something incorrect in the ad copy or something related to the landing page of the lower-ROI words that may be reducing the volume of sales and thus bringing down the average visitor value. You may also have an instance where the word cannot be paused, as it sells a critical product that has a higher lifetime average value than it does an average customer value. In these cases, the revenue is not in the first sale, but all the resulting sales.
Quality Score (QS) is a very important factor to track in your paid search campaigns, as it impacts the position and price you pay per click. Quality Score is determined by the engines dynamically at every search. However, Google provides some general feedback to Quality Score within AdWords as well. It is basically made up of the following elements, as listed at http://adwords.google.com/support/aw/bin/answer.py?hl=en&answer=10215:
The historical CTR of the keyword and the matched ad on Google
Your account history, which is measured by the CTR of all the ads and keywords in your account
The historical CTR of the display URLs in the ad group
The quality of your landing page
The relevance of the keyword to the ads in its ad group
The relevance of the keyword and the matched ad to the search query
Your account’s performance in the geographical region where the ad will be shown
Other relevance factors
If you are uncertain of the importance of this metric, consider the results of a detailed study run by ClickEquations on the impact of Quality Score (http://www.clickequations.com/blog/2009/03/the-economics-of-quality-score/). Table 4-9 shows how they found it impacted bids.
Table 4-9. Quality Score impact on cost per click for Google AdWords
|Quality Score||Discount or increase in CPC|
|10||Discount of about 30%|
|9||Discount of about 22.2%|
|8||Discount of about 12.5%|
|6||Increased by about 16.77%|
|5||Increased by about 40%|
|4||Increased by about 75%|
|3||Increased by about 133.3%|
|2||Increased by about 250%|
|1||Increased by about 600%|
By having an understanding of what you can expect in terms of discount or increase, you get an idea of how important Quality Score is to you now. It can be the difference between paying less than a dollar per click to more than six dollars per click. Monitoring your Quality Score and being cognizant of the factors Google lists will ultimately help you further drive down costs and improve your ROI by reducing the average cost per acquisition.
You can find your Quality Scores by navigating to your campaigns and then your keyword lists (Figure 4-6). In the “Status” column, you will see what looks like a word balloon; moving your mouse over this will show you the QS for each keyword set up in your campaign. Take note of the keywords that have a low QS, and consider pausing or modifying them to improve their QS.
Quality Score can be improved by targeting pages that are more relevant—try including the keyword in the page title, in the URL, or in H1 tags. Quality Score can also be improved by altering your bid strategy and improving the click-through rate of the words. Quality Score will also improve with historical trust. Over time, if you run campaigns with high CTRs targeting relevant pages, your QS can also improve.
The last topic to cover is seasonality and words. In both SEO and paid search, you may see some unusual spikes in certain terms during the year. This phenomenon is referred to as seasonality. It may be influenced by weather, holidays, events, or other annual occurrences. Seasonality is something that should be fairly predictable for your campaigns, as it is something that occurs annually. Do you have an annual announcement that goes out? If so, you may expect a spike after that announcement. Do you see sales increases around Valentine’s Day or some other holiday? Again, that sort of seasonality is predictable. To identify susceptible keywords, you can test your words in the Google Keyword Tool. Figure 4-7 is an example of a seasonality effect. Searches on the term “valentines” spike in February, with some occurring in January but virtually nothing through the rest of the year. We can also look at the term “engagement,” which you might expect to show some seasonality relating to Valentine’s Day. We see some increases on this term in March, April, August, and October, but there doesn’t appear to be any sort of pattern between “engagement” and “valentines” in terms of seasonality. If we ran a jewelry store, we might therefore decide it is not important to be aggressive on the term “engagement” in January and February, when many of our competitors may be out trying to get more traffic. However, if our sales tell us differently, we might decide we need to compete in that market regardless of what the search volumes tells us.
Being aware of seasonality can help you prepare campaign budgets, ads, and bids, possibly getting them to a point where you can turn them on and off. For your SEO campaigns, it allows you to plan in advance what terms you will need to have on your pages at different times of the year if you want to rank on some of those seasonal keywords. If you plan to rank on “valentines” in your SEO campaigns, you should start to work on building your ranking in late December, to ensure you have plenty of time to get indexed, and start to build links to the page through link-building campaigns. There should be no excuse for annual occurrences catching you by surprise. Being prepared for seasonal business spikes will mean you can spend more time optimizing and less time worrying about implementing.
Looking back at this chapter on words, we see how important it is to both understand and track how we perform on different keywords, as well as the impact tying the wrong words to the wrong page may have on our conversion rates. Poorly set-up ad groups can result in low Quality Scores, which means increases in your cost per click. SEO pages with many words clustered around them that are not tightly related to the goal of the content may also result in low conversions. Segmenting and refining your keywords will help you increase ROI. Although this may feel like a lot of work, it is a short-term pain for a long-term gain. We will come back to some more keyword issues when we get into tracking competitors. We will learn how to compare how our competitors performed on those words and how we can use analytics to figure out what is needed for us to surpass them in the results page.