Initial reactions to SSL Search were mixed

SSL search has been live for over a month now, and for some industries, the percentage os search terms ‘not provided’ keeps on going up. Ironically just as data driven decision making becomes more important to marketing professionals, one of the more valuable sources of information has become compromised.

Keywords are far more revealing about what a visitor is trying to do, what their interests are and how they describe products or ideas. Segmenting traffic by the words they used to arrive on the site can reveal a lot about how effective a campaign is through changes in campaign terms, identifying returning customers through navigational terms like the use of brands and URLs as queries, and also the success or otherwise of search optimisation campaigns.

Read all of Google’s secure search and missing data on MarketingMag.com.au

Rank first in Google! Or Last, maybe.

The other week Google introduced a new ad position on their results pages, at the bottom of the page. Reporting in AdWords will only display two ad positions, top and other, grouping impressions on ads at the bottom of the page with those generated from ads to the side. The ‘Top vs. Side’ segment became ‘Top vs. Other’.

In the blog post announcing the launch the change was explained as providing benefits for the user experience without compromising the advertisers’ interests:

In many cases, we have found that displaying ads below search results fits better into the user’s flow as they scan the page from top to bottom. On average, this placement performed better than side ads in terms of click-through rate in our tests.

This change was not made without testing and enough data to give a clear indication of what the likely effect on the market would be. Ads have been seen in this position in a number of different markets prior to the announcement. Pre-launch testing is not all that surprising from Google and has been seen many times in the past.

Movement in the SERPS

Is there any value to the advertiser in appearing at the bottom of the search engine results page (SERP)? It is almost certain that, now that it is launched, Google found no problem for themselves, however their interests do not always correspond to that of their users, or customers. Adding ad positions below the fold has even attracted the attention of non-digital marketers, and how this change and the uncertainty it introduces would affect the perceived value of an AdWords click.

Click throughs by Average Position on the first page

Graph from Click Throughs in the Search Results

A lot of data has been graphed and blog posts written on how organic listings below the first two do not receive a lot of traffic. The difference between the click through rate of AdWords ads positioned above the organic results and those to the right is significant. Putting ads below the tenth spot certainly doesn’t seem like it would increase the number of clicks each ad could get. But maybe that is not the point.

Splitting ad positions between the top and the bottom of the SERP differentiates the two positions further than between the top and the side. The top results will appear less cluttered with fewer links, even with the addition of more Google Places content.

For the AdWords advertiser, the perceived and probably actual value of a top three position also increases. With additional ads moved to the bottom of the page, the amount of above the fold screen real estate available is reduced.

Differentiated Impressions

Like the introduction of above the fold bidding on the content network, creating further differentiation between advertising inventory in AdWords creates additional products. Moving some of this inventory below the fold creates a greater difference in value, and potentially intensifies competition for one class of product.

The cost for placements above and below the fold will change as a consequence of this. The cost of traffic for placements consistently appearing below the fold will probably increase. Decreased visibility will affect their click through rate (CTR), and as a result negatively affect Quality Score (QS).

Placements at the top of the page will probably benefit from this. What one position loses in exposure the other should gain. The Quality Scores for the top placements should improve, if they receive a greater share of the advertising traffic for any given SERP. To what degree this will result in a lower cost per click will depend on what, if anything, changes in bidding behaviour from the other participants in the market.

The most likely outcome from Google’s latest change to how AdWords ads are displayed is an increase in revenue per impression. Creating greater differentiation between advertising inventory will encourage advertisers to modify their tactics and spending to maximise impressions above the fold, possibly raising the average cost per click. Moving one kind of ad position to a position below the fold will increase the value and yield on those that remain above, while the almost inevitable decline in CTR and QS will increase their cost per click.

Update: Google announced in a blog post titled New ways to take action on top of page bid estimates new tools to automate bidding to appear specifically above the fold. While it is not like the options introduced with above the fold bidding on the content network, the effect is the same.

And for now, this is correct. Currently the only keyword referral data being replaced with ‘(not provided)’ is that generated by users who:

• Search through Google.com,
• While signed into Google accounts, and
• Click on an organic listing (AdWords keyword referrals are still passed).

Internationally the volume of organic search traffic that meets these criteria is very small, and does not account for much USA traffic either. There is very little available data at this point, and most of these assumptions are based on what people are seeing on their own sites (now that it is it is possible to start to measure the impact) and Matt Cutts’ informal estimate of the percentage of searches performed by signed-in users.

Important, not Urgent

Pay for traffic and get the data for free

The issue is not how much organic search traffic is ‘(not provided)’ now, but what this number might be in the future. SSL Search isn’t an urgent issue, just an important one. Google builds systems for scale, and has stated their intention to take this further:

As part of our commitment to provide a more secure online experience, today we announced that SSL Search on https://www.google.com will become the default experience for signed in users on google.com.

Making SSL Search the default for signed in users won’t be confined to just Google.com forever, and the number of users who remain logged in while using the internet won’t remain the same either. Most of Google’s popular services like Gmail, Google Docs and, optimistically, Google+ work better when you are signed in. Over time there will be more people searching while logged in, not fewer, and SSL Search is certain to be rolled out across more Google TLDs.

Won’t Someone Think of the Users?

It should not be a surprise that Google is working on making a more secure search experience more widely available. Google pays attention to Electronic Frontier Foundation (EFF) and takes the user experience of search seriously. This culture is often reflected by the employees; for example, when SSL search was first introduced by Google in May, 2010, before being moved to https://encrypted.google.com/ in June, Matt Cutts posted A few thoughts on SSL Search on his personal blog on its introduction:

I believe encrypted search is an important option for Google searchers. The Electronic Frontier Foundation (EFF) has asked for secure search in the past (see this post from 2009), and I credit them for helping to put this on Google’s radar. Another inspiration that helped to spark this project was Cory Doctorow’s book “Little Brother.” It was one of my favorite books of 2008 and while I won’t go into the book’s plot here, it’s a quick, fun read. “Little Brother” also makes a compelling case for encrypting HTTP traffic on the web.

Recently, a few ISPs in the USA demonstrated the value of encrypting search referrals for users by redirecting their customers’ search traffic based on the keywords they used to content the ISP selected. In fact, it is the value that SSL Search has for the user that means that Google is providing organisations such as schools a NoSSLSearch option.

Google’s commitment to their users’ privacy and HTTPS Everywhere means that their latest rollout of SSL Search for logged in users is a logical progression, securing their users’ search queries and protecting their privacy. Unless they click on an AdWords listing. If you are paying for clicks, those keyword referrals are still passed through to the advertiser.

Organic Versus Paid Traffic and Data

Despite the Internet’s tendency towards paranoia when reading the motives of large companies, there is actually a good reason why referrals don’t get passed from HTTPS to HTTP sites. As Thomas P from the Webmaster Central Help Forums pointed out:

Clients SHOULD NOT include a Referrer header field in a (non-secure) HTTP request if the referring page was transferred with a secure protocol. (Source: Hypertext Transfer Protocol — HTTP/1.1)

Following these standards means that sites behind HTTPS would still see keyword referrals. But this is not the case with SSL Search; what Google is doing is a little different. In a comment by Eric Wu on Google+, he suggested that Google is not using HTTPS as you would expect (follow the link, his explanation is well worth reading). Searching via https://google.com/ differs from the standards (unlike https://encrypted.google.com/, which still follows them) in two important ways:

• Organic referrals won’t be passed to sites entirely behind HTTPS
• AdWords referrals are passed to sites on HTTP

Both of these are odd behaviours, and seem at odds with the aim of protecting users’ privacy. Passing AdWords keyword referrer information to advertisers is not all that surprising once you think of Google as a business. Danny Sullivan pointed out on Search Engine Land in his Google Puts A Price On Privacy post that provoking a negative response from the people who made Google nearly $10 billion last quarter would not be good for business.

Information is valuable

As long as Google does not increase their number of users...

It is the loss of data due to SSL Search that concerns most online marketing professionals. Search referrals often indicate changes in how queries are structured, shifts in brand awareness and use, the effectiveness in offline campaigns based on certain phrases or words and give important insights into how language is used to describe products or ideas. Keyword referrer information is important for conversion optimisation and can inform ongoing IA development. As limited as the scope of Google’s initial implementation of SSL Search is, the fear that it will account for a greater share of search as time goes by is legitimate, and has real business implications.

In a number of blog posts Google has suggested that Webmaster Tools data can replace the information lost in Google Analytics, with Webmaster tools query data currently available in Google Analytics. However the information provided by Webmaster Tools is substantially different to the keyword data in Google Analytics. It is restricted to only 1,000 queries, and is reported in general terms. It is useless for assessing search queries of more than two words for any site that receives a decent amount of search traffic. Google Webmaster Tools gives an overview of what is happening within Google, but in its current form, it can not replace Keyword referral information.

This might not actually be a bad thing for Google. Information, especially information on user behaviour, is valuable. Google will still be collecting search data from those using SSL Search; the change simply means that they won’t be sharing anymore. At least not without being paid first.

Balancing Privacy and Business

Google has taken a compromise position on user privacy with this implementation of SSL Search. Even as they recognise the value of HTTPS Everywhere for users, Google has acknowledged that referral data has value for their customers (their advertisers, that is – those that don’t pay money are not customers). The result is a feature that does not really protect users’ privacy or encourage more websites receiving both free and paid Google traffic to implement HTTPS.

It is interesting that Google’s other SSL Search URL, https://encrypted.google.com/, handles referrers differently to https://google.com/. With clicks on HTTP URLs, https://google.com passes a HTTP referrer, stripped of keywords, but https://encrypted.google.com does not. When the destination is a HTTPS URL, https://encrypted.google.com passes referrer information as you would expect, but https://google.com does not, unless it is an AdWords link. There is a more complete post, with a chart, at Google Modifies SSL Behavior — and the Results Are Troubling.

To a cynical outsider, it would seem that the only objective Google has actually achieved with SSL Search is protecting their advertisers’ keyword referrals while earning favourable press for defending the privacy of innocent Google users from those nasty, evil, SEO scammers.

Generic versus Brand Searches

Creating a Generic versus Brand Search Rule

Brand search term rules are so important that Google Analytics uses ‘Generic Keywords vs Brand Keywords’ as an example name during the custom Multi-Channel Funnel grouping creation process. It is a very powerful tool, assessing the general performance of a search related campaign, paid or otherwise. Further segmenting search by paid and unpaid and adding both direct and referral traffic makes it even more useful.

Search, Direct and Referral Paths

Search marketing is a great opportunity to reach people without much awareness of your brand and change their behaviour. Online search is more or less a de facto collection of bookmarks, where people enter queries based on URLs, brand or product for sites they have already visited. Tracking queries by the presence of brand terms through Multi-Channel Funnels can reveal how often non-branded search leads to branded queries, direct or referral sales. This can be a great tool for selling the value of SEO internally by demonstrating its value in the customer’s product research behaviour.

Regular Expressions for the rules

Regular expressions are a great way to keep your sanity when you are building the rules for any group of keywords. For example, brand traffic is not just going to be attributed to the correct, proper name of the brand. Variations in spelling and construction have to be accounted for. As an example, for tracking queries on the brand ‘Greyhound Australia’, the regular expression would include the brand and its most common misspellings:

greyhound|grey hound|grayhound|gray hound

The same kind of regular expressions can be used in other rules, like product terms or search terms targeted in a focused SEO project.

Product Terms with or without Brand

Whether or not a search contains a brand term only really speaks to awareness. Brand terms alone don’t distinguish between navigational, informational or transactional queries. A search for just ‘Greyhound Australia’ is clearly navigational. The search lacks any other qualifying term and Google’s recent change to how sitelinks are displayed for some queries seems to indicate even they regard such searches as purely navigational.

Brand rules with and without product terms

Creating rules that label traffic using qualifying terms and traffic that doesn’t can make the difference between navigational and informational searches in the funnel clear. Differentiating between a search for ‘Greyhound Sydney to Canberra’ and ‘Greyhound Australia’ is the difference between seeing a funnel with two brand searches, and a funnel with a search for a product including a brand term and a navigational brand search.

Campaign Source Attribution

There are more online customer acquisition channels than search. Directories, display networks, guest posts, links and advertorial are all tools that can reach new customers or initiate research and purchase behaviour. Multi-Channel Funnels and custom groupings make it easy to see how they feed into other traffic sources.

Tracking more than search traffic

Adding rules for campaign URLs used in advertising or branded content and source URLs such as social media touchpoints make it possible to see how these activities drive customer behaviour. The ability to identify campaigns that drive sales both directly and through other channels make informed decisions regarding future activity easier.

Custom Groupings: Just Start Using Them

Multi-Channel Funnels make it easier to see what works, what doesn’t and how certain activities can lead to sales. The ability to differentiate between broad information searches and brand focused queries and identify how they are linked across different search mediums is valuable. Tracking conversions across campaigns and channels makes it easier to make informed decisions about social media activity, guest posts and display advertising or directory lists as well.

Plan how to integrate Multi-Channel Funnels with your current online activities. The process to get started can be as simple as:

• Create custom Channel groupings focusing on sources that matter:
  • To you
  • To the people you report to
• Create new Channel groupings from existing ones to monitor new campaigns and sources
• See what works, see what doesn’t, see what can be improved

What are your Must Have Multi-Channel Funnels?

The single most important thing I think Multi-Channel Funnels can give you is the ability to see what broad research behaviours lead to direct navigational activity, either as direct visits or navigational brand searches.

How are you going to use these new tools in Google Analytics?

Music in the cloud, and buying Skype

A Distributed Microsoft Social Network

Microsoft has a successful email service, a popular gaming platform with an online social network and a messaging client with a large install base. Also at the end of last year, Microsoft released a Facebook app for Live Messenger and began to incorporate social feeds into their own mobile and desktop chat clients. Microsoft already has a network covering multiple platforms and devices. Skype is an expansion of this network, and with Qik, it certainly shows some potential for mobile.

A more social Live

Reaching the Multi-Screen, Multi-Device, Multi-Platform User

There is more than one way to reach a new destination on the Internet. There is also more than one kind of software to use, device to operate and method to find new things. An Internet where banners and search advertising displayed in a browser are the only means of advertising no longer exists.

Mobile, gaming and social platforms are as important for building awareness as online advertising targeting desktop computer usage. The Microsoft Advertising blog has published a lot of content on reaching the multiscreen consumers, and advertising across multiple devices.

Buying An Audience

A number of interesting stats on Skype users were shared after the announcement. The most impressive was Skype’s current 170 million users and 40% year over year growth.

Buying Skype (with Qik in tow) gives Microsoft more users, much in the same way that their deal with Nokia will increase their Windows Phone 7 install base from 1.4% in Q4 2010. Even if Nokia loses an expected market of 50 million smartphones in 2011, the market share left by the time a Windows Phone 7 handset arrives will still be more than Microsoft has now.

Attention is all online

With information as easy to create and publish as it is now, it is curation that matters. Curation is more than a directory and a search engine now. Microsoft’s range of services, software and devices position them for a post-PC world, with its fragmented user experience. Microsoft’s own published reports say that we now:

…start our sessions in what I would call our intimate zones seeking personal information and contact through email, social networks, blogs etc.

Microsoft is building a distributed social network, covering computers, gaming, mobiles and tablets. They are creating a platform that can include Hotmail users, Xbox Live users, Facebook users, Windows Live users and now, Skype and Qik users. It reaches consoles, computers and mobiles and is accessible from different devices and platforms. It just isn’t limited to a single domain.

The types of content don’t really change from one device to the next, but how I experience, find or explore content, which software, applications, platforms I use, and where and how I connect do. The people I am connected to, how and where I search, the connection stability and speed and the specifics of the device I use all affect how and what I consume online.

There is more than one way to find or do things online. Publication is easier than ever and the Internet has an almost unlimited capacity for content. TV and radio only ever had 24 hours per day to fill, newspapers have a set number of pages, and the market could only sustain a limited number of these entities. Online, these limitations don’t exist; there is always space to publish just about anything. If the content created is good enough for the user, it is attention that matters, not the platform.

Some of the best business models online involve providing a platform (Facebook, Apple App Store), a search engine (Google, YouTube) or some form of aggregator (Flipbook, Netflix). Companies like DemandMedia are the exceptions that prove the rule. Their content production model works, both before and after Google’s Panda Update (an algorithm change targeting ‘low quality’ content, which may or may not be ‘content farms’), because it is integral to their attention and audience building strategy.

What shapes the user's experience of the Internet

Many of the things that influence a user’s online experience, how they search and explore, and the information they consume fit into one of these broad categories:

• Infrastructure
• Connection Speeds and Stability
• ISP
• Application & Address Layer
• Devices
• Software
• Navigation & Discovery

A user’s experience of the Internet is shaped as much by these as by many other factors: how they find new content, the software they use, the device used, how it connects to the internet, the ISP. There are also many other services running in the background, just out of sight, such as the DNS provider, application layer and physical hardware, which also affect the end user experience. Their impact can be as dramatic as an Armenian woman cutting two optic-fibre cables or as subtle as connection speed and stability.

Sites and Portals, Clients and Servers

There is a brillant image on Doc Searls Weblog, from the post A sense of bewronging. It is a photo of a cow and a suckling calf. Doc Searls used it in a slide deck to illustrate the relationship between users and sites on the commercial web:

It’s a calf-cow model. As calves, we request pages and other files from servers, usually getting cookie ingredients mixed in, so the cow can remember where we were the last time we suckled, and also give us better services.

Online search, commercial websites and social and advertising networks now track user behaviour and information better than ever and adjust their content to suit. This trend is neither new nor limited to Google and Bing’s forays into personalised search.

Arguably this is not in the best interest of the users, especially in search. Social signals might be handy for finding a restaurant, but their value declines as the information falls further and further outside of the social network’s aggregated sphere of competence. For example, the sites a creationist might like would be utterly pointless for a query on evolutionary biology, even if the text includes the same terms.

These one-sided relationships are not just limited to the commercial web, but extend to hardware, software, the ISPs we use and ultimately to the digital and physical infrastructure of the Internet. Individual users are being moved away from an objective Internet. Tools like Google, Bing, Twitter, News.me, Facebook, and any other platform or service that serves content person by person are creating a siloed, Narrowcast Internet.

The Why and How

Search is credited with being THE way people find information online. Algorithmically generated lists of links, paid or otherwise, account for a significant amount of traffic on the Internet. Despite indexes full of optimised commercial content and with declining search literacy (Trust Online: Young Adults’ Evaluation of Web Content), arguably no motive for the search engines to provide more than An Adequate Search Result, search is still seen as the best way to find new information.

Search Engine Result Pages (SERP) are not the only place people spend their time or discover new stuff. Hitwise’s ‘Top 20 Sites & Engines’ report for the US datacentre indicates a more nuanced picture of online behaviour. The top five most visited sites for the week ending 23rd April 2010 were:

1. Facebook
2. Google
3. Youtube
4. Yahoo! Mail
5. Yahoo!

The top five for Hitwise’s ‘Top 20 Sites & Engines’ (week ending 23rd April 2010) report from their Australia datacentre was not all that different:

1. Google
2. Facebook
3. Youtube
4. Windows Live Mail
5. ninemsn

Portals, social media and email sites matter as well as search, either web or video. Other discovery modes such as link sharing via social media, email and outbound links on articles are as important for generating attention and an audience as a listing on a Search Engine Results Page (SERP). The post, New Microsoft Advertising Study on “Living with the Internet”: What’s driving web behaviour? on the Microsoft Advertising Blog makes the point based on their research that users:

…start our sessions in what I would call our intimate zones seeking personal information and contact through email, social networks, blogs etc.

Curation matters, especially online, and curating requires choice. Linking, displaying, referring to content inadvertently prejudices the user against what they don’t see. The link a friend would tweet or email you to answer a question might not be the first one you would find on a Google results page. The product description you see in a company’s iPhone app can vary from the one on their Facebook page. Site content and ads can change depending on the device being used too. What a user finds to answer a question or complete a task does change depending on how they find it.

Search and curation

There are a number of different ways a user can navigate from one place to another online. Non-digital media directs users to new information, so do kinds of feeds delivered through single or multiple source applications like News.me, last.fm and The Australian’s app. Most navigation modes fall into one of the following groups:

• Search Algorithmic Link
• Search Paid Link
• Shared Link/Bookmark Link
• Non-Search Paid Link
• Unpaid Link
• Personally Bookmarked
• Direct Feed API
• User Entered Destination

Motives, Devices and Further Fragmentation

User intent plays a huge role in determining which tools they use and how. Microsoft’s ‘Living with the Internet’ report identified six different motives for the majority of online activity:

• Communication
• Information
• Entertainment
• Surfing
• Transaction
• Creation

The report explored the different motives behind Internet use by PC, laptop and netbook compared to smartphone and tablet devices. The smartphone and tablet users from the report focused on fewer motives per session than those using laptops and desktops. Mobile and tablet users were also less likely to cite entertainment and transaction motives for their time spent online.

Differences in the device’s interface, screen size and the user’s concurrent activities all probably contributed to this pattern, at least in the population studied. Ease of managing multiple browser tabs and applications probably plays a role in differentiating behaviour from one device to another. Keyboard versus touch or phone keypad input is another factor.

Mobile Matters

A post called ‘Some Milestones We Will See This Year in Mobile Statistics’ on the Communities Dominate Brands blog bought up some interesting figures on the 4.6 billion actual mobile phone handsets in use last year, along with a few interesting predictions for the coming year.

Last year, of the 4.6 billion mobile phones in use:

• 96% of all phones in use worldwide have at least a basic browser
• 71% of all phones in use had a ‘real’ web browser that was HTML compatible
• 59% can do basic apps via Java or Brew

Mobile devices are becoming an even more significant part of the user’s experience of the Internet. If all you have to use the internet with is a phone, you can’t play EVE online, WOW or run Steam. If you only use a desktop, you don’t get to use augmented reality. Use a dumb phone, and chances are you are restricted to WAP or a seriously impaired experience on non-mobile optimised sites. However, if you are using a smart or feature phone, you can access a different application ecosystem, and use location and augmented reality tools.

Cows all the way down

Our virtual world is a network of inter-dependent nodes hostage to Armenian women cutting cords, governments building walls or flicking switches, unreliable networks, hosting companies cutting cords, domain names taken down, and so much more.

A user’s experience and the information they find is the result of a large number of factors. Online services are diversifying and use more signals and cues for sorting and curating information, creating billions of different user experiences. Each of these experiences is a result of the many layers of dependencies and gatekeepers between the user and the rest of the Internet, from the hardware they use to how the information is placed in front of them.

To borrow that mental image from Doc Searls, it’s cows all the way down.

Total Impressions by Quality Score

The data used in the last blog post demonstrated that Quality Scores of between 7 and 10 featured prominently in Average Positions of 3.9 and higher, and that the bulk of impressions the campaign received were a result of terms with a Quality Score of 7 or higher, barring one outlier at Quality Score 4 in the dataset.

Impressions for Quality Score or Impressions for Keywords

That most Impressions occurred through a Quality Score 7 or higher keyword becomes less remarkable once the number of keywords at each Quality Score is considered. An easy majority of the keywords from the data used in the original blog post had a Quality Score of 7 or higher.

Number of Keywords by Quality Score

How do the keywords with a Quality Score of 7 differ from those at other levels? Google has publicly stated that Quality Score is calculated using a number of factors like:

• The historical clickthrough rate (CTR) of the keyword and the matched ad on Google; note that CTR on the Google Network only ever impacts Quality Score on the Google Network — not 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

Ignoring the “don’t be an idiot” factors such as landing page quality and relevance of the keyword to the ads in the ad group, Quality Score is mostly influenced by the performance of the search terms, the keyword, the ad and your own account history.

Impressions, Quality Score, Click Through Rate and Cost Per Click for Quality Score 7

Click Through Rate, Exact versus All Shown by Impressions for Quality Score 7

Quality Score is strongly associated with the Click Through Rate (CTR) on the exact match of the keyword in the campaign. It is not easy to get to this data through AdWords, either online or with their desktop tool, and usually requires a little work in a spreadsheet.

The dataset use for an Impure.com workspace compares the keyword’s exact match Impressions, CTR and Cost Per Click (CPC) with those listed for all searches it accounts for. The exact match data was acquired through the all search term report available through the AdWords account management site.

For keywords with a Quality Score of 7, there is a relationship between the exact match Impressions it receives and the CTR it has. Keywords at this Quality Score with more impressions tend to have a lower Click Through Rate than those with fewer. This pattern does not hold for the Click Through Rate of all impressions triggered by the keyword.

Exact Match CPC by Impressions and Click Through Rate for Quality Score 7

The relationship between exact match Impressions and CTR also exists for Cost Per Click. Keywords with more Impressions also pay less per click. Again, once you use data for all Impressions triggered by the keyword, this relationship seems to disappear.

All Searches CPC by Impressions and Click Through Rate for Quality Score 7

For the campaign examined, it appears that it is the volume of Impressions that the keyword receives for a Quality Score of 7 that can influence CPC as much as Impressions. However, the relationship this has with the cost for all clicks received as a result of phrase or broad match is a little unclear.

Regarding Quality Score as a bulk discount rather than a relevance score is a more meaningful perspective. At least within the dataset used and for a Quality Score of 7, the more Impressions and the better the Click Through Rate, the cheaper the traffic.

Defining Click Through Rate, Average Position and Quality Score

The definitions for Click Through Rate, Average Position and Quality Score below have been quoted from Google AdWords help pages.

• Click Through Rate (CTR)
  • The number of clicks your ad receives divided by the number of times your ad is shown (impressions). (Source)
• Average Position (Avg. Pos.)
  • Refers to the average position on a search result page that an ad appears in when it’s triggered by that keyword. (Source)
• Quality Score
  • Quality Scores help ensure that only the most relevant ads appear to users on Google and the Google Network. (Source)

Visualising Click Through Rate by Quality Score & Average Position

A guide to the Impure.com graph linked to. Either use the embedded graph, or click this image to go to the active work space and explore it further.

In this example set of campaign data the distribution of high CTR by both Quality Score and Avg. Pos. is clustered near an average position of one, and a Quality Score of between 7 and 10. A lot of the traffic in the campaign had a Quality Score of 7 and as a result, activity associated with that Quality Score uis present in most ad positions above the 5th spot on the result page. The workspace was created by Impure.com, a rather cool tool that makes it possible to share visualisations and let others explore them further.

While a high Quality Score tends to mean that the keyword is in a higher position in the Search Engine Results Page (SERP), it is not the most significant factor. In the attached graph Click Through is more closely linked to Average Position, regardless of its apparent Quality Score.

Click here for the full, active, workspace.

The graph shows average CTR for an Average Position and Quality Score. Position one and Quality Score ten are in the uppermost corner, with the largest spike in both data-sets. When the 3D Surface visualisation first opens, Quality Score is displayed as ten on the rightmost line, through to one on the leftmost. Average Position is depicted as highest on the page closer to the original point of view.

The Graph and the Data

The first file linked to via the switch, ‘all-qs-avepos.csv’, uses a complete campaign, whilst the second, ‘number-ctr.csv’, displays only a small sample from the original. The third, ’1-39-ctr.csv’, only has data for average positions 1 through to 3.9.

In the campaign examined, the most active average positions are between 1 and 3.9. Whilst there are impressions recorded for other positions on the page, it is this area where most of the activity is recorded. A histogram where you can explore how impressions, clicks and so on related to reported average position can be found here.

Determining Average Position in Google’s Results

Cost per Click, Average Position and Click Through Rate

There is a strong connection between an ad’s Average Position and its Click Through Rate in Google’s search results, in the same way as there is a strong relationship between Quality Score and Click Through Rate, even with the indeterminate number of other ranking factors Google AdWords takes into account.

A Quora question; “Does “broad match” on Google AdWords often lead to lower quality scores for keywords?” touches on a few of the more interesting points regarding just how Google uses Click Through Rate data to determine a keyword’s Quality Score. Google’s own explanation of the topic is also worth reading through.

Ultimately, if you are willing to spend enough, you can get almost anything to appear near the top of Google through AdWords. Quality Score is in effect a discount for both providing an ad that does not seem to annoy the users and bulk buying traffic, represented by a higher Click Through Rate. Quality Score deserves its own blog post, and so it will be covered further in another post.

But do not mention the content farms

This February has been a month of drama for the search industry. It started with Google accusing Bing of cheating at a search panel at the start of the month. Google followed up the lively panel discussion by publishing a lot of screenshots explaining why they are convinced that Bing was using their search results as a ranking factor.

The write-up on how they did this is pretty comprehensive, but just briefly: they planted artificial queries and set sites to rank for them within their own live directory. They then used the queries and clicked on the links they planted there with computers running Windows, Internet Explorer and with the Bing toolbar installed. It was published on Google’s official blog after the talk, and found its way into the USA’s mainstream media, in a severely abridged form. Bing responded shortly after as well, and put their case forward on their own blog.

In the discussion that developed around this very public fight, kilobytes of text were used to list which search engine came up with what. Google’s copying of Bing’s left column and infinite image search scroll were mentioned, as was Google’s universal search. Issues like whose content Google was incorporating into their own product, sometimes against the owner’s wishes, emerged too. Google’s original statements regarding their own manual control over a site’s position in search emerged, and their original statement “we do not have the capability to manually change it” was brought up.

In all the excitement however everyone seemed to forget what the search panel was expected to be about: content farms, and poor quality content funded by AdSense. The discussion about search quality that permeated the search and Internet marketing community seemed to be forgotten. Even the long running debate over whether or not Google was favouring their own products, such as YouTube and Google Places, was forgotten for a little while.

Adapted from BrisbaneOnlineMarketingMeetup.com – Purported SocMed / Digital Marketing Gurus

Google is on record saying:

“Twenty percent of the queries Google receives each day are ones we haven’t seen in at least 90 days.”

Broad and phrase match makes it easy for an advertiser to deal with this and keep the size of keyword lists under control. Both of these match types can also produce irrelevant traffic. The terms added to the campaign may be similar or frequently associated with the keywords in normal search. This does not mean they are always relevant.

Irrelevant queries are not the only reason traffic might be unprofitable. Issues such as competition, pricing, geography or other circumstance can render some queries too expensive to be profitable, or even accessible for your campaigns.

Setting negative terms is the easiest way to reduce a campaign’s exposure to bad traffic. A negative term prevents an ad from showing for search queries containing it. In campaigns using terms relevant to multiple industries, negatives make a difference to the cost per acquisition.

Finding the Z for X and Y

Finding negative terms for a campaign is an interesting exercise. Unproductive clusters of terms can be found using any tool that allows for filtering by words and analysis of relevant metrics.

Simple hierarchical clustering dendrogram

For example: A campaign is over target CPA, and features two keywords, ‘A’ and ‘B’. Both keywords are using broad match. Historically these terms have generated sales and a consistent level of traffic. A search term report showing all search terms for keywords ‘A’ or ‘B’ reveals a number of other commonly occuring terms for which the ads display. Each of these terms above a certain threshold can then be analysed.

A quick analysis of other terms that account for a certain percentage of impressions (or other metric of choice) has revealed that traffic with the term ‘Z’ frequently exceeds the maximum allowed CPA, whilst other significant terms like ‘X’ and ‘Y’ are on target. So what now? Should it produce minimal sales, it should become a negative term. If it does produce some sales, the traffic for the term can be analysed again, to identify if there is another term that should be removed from the campaign.

As the number of significant terms in a campaign increases, so does the complexity of the analysis. In the example attached, the campaign’s targeted CPA on reporting period has been revised, and it needs to be adjusted to match the new target.

Campaign Term Nodes at higher than target CPA

The dataset used in the example contains a range of descriptive terms that relate to locations, product description, usage and alternative products. In this example, hierarchical clustering reveals a few terms that appear across multiple nodes. Each node is split until it reaches a threshold, after which the data would cease to be statistically significant.

The Intersection of Clusters

What you are looking for is the groupings that have the greatest impact on the metrics the campaign is measured on. If it is return on investment, the CPA is used for assessing the clusters of terms; if it is time on site, or even raw volume, those would be used instead. While this can be done within Excel, across multiple worksheets, the information is best dealt with as a network, and visualised as such.

And how to get it all wrong

There is one huge error that can lead to otherwise productive terms being discarded: decisions made without enough data, either as a result of a small dataset, or overfitting of terms. Search Engine Marketing traffic is very prone to volatility at low volumes, and the problem becomes more pronounced the lower the chance of a sale per click is at the ad.