New data on the click-through rates (CTR) from different positions within natural search results (SERPs) is always accompanied by a flurry of blog posts and comments from the SEO community. When AOL released data on the click-through rates from SERPs by their customers it was instantly seized upon by several bloggers (e.g. 1, 2). More recently, when Wunderman presented click-through data at an Ad-Tech event (pdf) a blog post presenting the data attracted 26 comments. This, perhaps, isn’t surprising since such data can form the basis for estimating the commercial value to a client of moving up to the top ten, top three or even top result on the search engines.  What is perhaps more surprising is that there is no definitive and accepted data on CTR from different positions on search results pages. Of the 26 comments on the Wunderman data, mentioned above, 15 of them questioned or challenged the data, its source, its validity and/or its value.

Meta-analysis

One approach to deriving definitive data on a topic is meta-analysis. The principle behind meta-analysis is that ‘truth usually lies amidst many approximations’. Re-analysing data from several sources enables trends to be identified with greater accuracy (the effect of outliers in any single data set is minimised) and with greater confidence (it is based on several sets of independently collected data).

Five data sets on click-through rates from natural search were found through web-based research.

1. Andrew Gerhart (Google Analytics data; click-throughs=6759; date=2009; source data)
2. AOL (Transaction log data; click-throughs=4,926,623; date=2006; source data)
3. Granka et al (Experimental study; click-throughs=275: date=2004; source data pdf Fig 1)
4. Pan et al (Experimental study; click-throughs=200: date=2006; source data Fig 3, ‘normal condition’)
5. Wunderman (transaction log data?; click-throughs=?; date=2007; source data pdf page 6)

The data from these sources was collated, normalised (the percentages were re-calculated for only the top 10 natural search results) and then analysed using the SPSS statistics application.

An advantage of these data sets (fortuitously!) is their diversity: two are from transaction logs, two are research data and one is from Google Analytics. They also range in sample size from a couple of hundred to several million. Consistencies found in such diverse data are likely to have wide applicability.

Results

The graph below shows the percentage of click-throughs from the top 10 natural search results, from the five data sets, and also the best-fit regression line through the data.

Effect of search results position on click-through rate

Regression analysis gives a best-fit equation of CTR=0.59 x Position^-1.94 (R²=0.96 p<0.001).

Interpretation and use

Given the diversity of data sources, there is a remarkable level of consistency within the data. It is not possible to give meaningful measures of variance because each data point is itself an average with unknown variance. Even without this, however, we are able to predict that moving a search result from position 10 to position 3 will increase click-through volumes by and average of 10-fold and moving from position 3 to position 1 will increase click-through volumes by 8-fold. Combining this model with data on search volumes allows us to estimate actual visit numbers. So, for example, given that there were 450,000 searches for ‘digital camera’ on Google in the UK during June, we can estimate that a site currently appearing at position 5 on the search results page could attract an extra 57,500 visits per month if they moved up to position 2 (search volume data from Google Keyword Tool).

Until additional data emerges, it is hoped that this model can form a basis for industry-wide agreement on the general relationship between position in search results and click-through rates.

3rd in a series of 3 posts - in the first post on singular vs plural keywords, I presented evidence to show that searches for singular and plural keywords changes the top ten search results on Google and therefore optimising for both is important. In the subsequent post I showed that whilst there are more than twice as many searches for plural keywords than singular, the opposite effect is found for particular keywords.

In this post the research data collected for the previous analyses will be used to estimate the opportunities available through search engine optimisation of singular and plural keywords.

Let’s consider 4 different SEO strategies regarding singular and plural keywords

1. All singular: optimise the singular version of all keywords - this is quite a likely strategy for anyone who believes that search engines will simply ignore differences between singular and plural keywords;
2. All plural: optimise the plural version of all keywords - this is the strategy likely to have been adopted by people who have seen one of the many reports (example) suggesting that plural keywords are searched for more frequently than singular keywords;
3. Category-specific singulars/plurals: optimise for either singular or plural keywords depending on which predominate within specific categories - e.g. laptop(s), mobile phone(s);
4. Optimised for both singular and plural keywords - for present purposes I will ignore the compromises this may involve and the adverse consequences for other SEO objectives that might ensue.

To make the opportunity analysis feasible and not too assumption-laden, let us simply explore the impact of these 4 strategies on search volumes, using the data collected for the previous post (see research methods). This provided search volumes on 1019 keyphrases that included the singular and plural versions of 7 seed keywords (bed, book, digital camera, laptop, lcd tv, mobile phone and sock).  If fully optimised for both singular and plural keywords (strategy 4), these 1019 keyphrases have a total search volume of 10.6 million searches per month.

Efficiency of different SEO strategies on singular and plural keywords

Optimising for all singular keywords risks losing 64% of the search volume. Optimising for all plurals or on a category-specific basis risks losing 36% and 31% of search volume respectively.  So, for a site that is search engine optimised entirely for singular keywords, optimising for both singular AND plural keywords has the potential to double search traffic.

This is the second post in a series of three on singular versus plural keywords - see 1. Singular vs plural: does it matter, 3. Singular vs plural: the opportunity.

We saw in the last post that singular and plural versions of keywords give different search results - optimising for both, therefore, does matter. In this post we start to answer the question ‘how much does it matter’ by considering the search volumes for singular and plural keywords.

Research method

The search volumes for singular and plural versions of keywords was investigated using Google’s Keyword Tool. Seven keywords were selected for this research: bed(s), book(s), digital camera(s), laptop(s), lcd tv(s), mobile phone(s) and sock(s). Five of these were selected because they were the keywords used in Robin Goad’s previous research on singular and plural keywords, which concluded that ‘plural terms are better at sending traffic to retailers than singular terms’. These keywords were: book(s), digital camera(s), laptop(s), lcd tv(s), mobile phone(s). Bed(s) was what started my interest in singular and plural keywords - some data I had from a client project raised some questions I couldn’t answer. And socks was added as a reference keyword: I wanted to explore what the data revealed about keywords with a strong bias in favour of either plurals or singulars - who searches for a single sock? For each of these seven keywords, the singular and plural versions were typed into Google’s Keyword Tool, with the ‘results tailored to’ English, United Kingdom, the ‘match type’ set to exact: the search volumes were recorded from the column titled Local Search Volumes April. The keyword tool returned the top 150 keyphrases containing either the singular or plural version of the target keyword (except laptop(s) where only 119 keyphrases were returned). These were then analysed for:

1. the number of keyphrases containing the singular or plural versions of the target keyword
2. the number of searches conducted for these singular- or plural-version keyphrases
3. the number of searches conducted for the singular- or plural-version keyphrases where both versions of the keyphrase appeared within the top 150 keyphrases (head-to-head comparison)
4. the number of searches conducted for the singular- or plural-version keyword on its own.

Results

Primed with the singular and plural version of 7 different keywords, Google’s Keyword Tool returned 1019 suggested keyphrases. Of these, 509 contained the plural version of the keyword and 510 contained the singular version - as close as it is possible to get to a 50:50 split with an odd number of results! Without the sock(s) keyphrases, which were expected to be strongly biased in favour of plurals, 500 of the top keyphrases were singular (58%) and 369 were plural (42%).  Google’s Keyword Tool reported that 10.6 million searches were conducted for the 1019 keyphrases. Of these searches (again excluding socks), 37% contained the singular-version of the keyword and 63% the plural-version. So, the overall findings from this research reveal that the top keyphrases contain 35% more singular than plural keywords but there is 70% more search volume for the plural keywords. Combining these two statistics to give search volume per keyphrase, there are 130% more searches for each plural keyphrase than there are for each singular keyphrase.

These overall findings, however, conceal some important exceptions. When we look at the number of keyphrases returned by Google’s Keyword Tool for each of our 7 target keywords we see that some are predominantly singular and others are predominantly plural.

Number of keyphrases using singular and plural keywords

There are more singular keyphrases for digital camera(s), lcd tv(s) and mobile phone(s) - bed(s) and laptop(s) have more plural keyphrases, but only just (57% and 54% respectively). Looking at the number of searches changes the predominance of singular and plurals yet again.

Number of searches for singular and plural keyphrases

Whilst 76% of the digital camera(s) keyphrases were singular, 66% of the searches were plural. The pattern for laptop(s) is similar but more modest (54% of keyphrases were plural but 54% of the searches were singular). Generalising patterns of singular and plural keyword use across different types of searches is, therefore, dangerous: searchers change their use of singular and plural keywords according to what they are searching for.

Making sense of how searchers use singular and plural keywords

Looking in detail at the differences between specific keyphrases that are mostly plural and those that are mostly singular suggests that there are 3 types of influences on searchers’ choices of singular or plural - factors I have labelled domain factors, linguistic factors and psycho-linguistic factors.

Domain factors are characteristics of particular domains of knowledge that pre-dispose searchers to use singular or plural keywords. In all the examples discussed here, the domains of knowledge are specific types of products - and it doesn’t take a genius to realise that searches for ’socks’ will substantially outnumber searches for ’sock’ because socks are sold in pairs. The data on sock-related keyphrases confirms this: 93% of keyphrases returned by the Google Keyword Tool and 94% of the search volume for those keyphrases was plural (i.e. socks as opposed to sock). There was only one major exception - searches for sock shop, which is a well known brand.  Similarly, the fact that searches for ‘books’ greatly outnumbered searches for ‘book’ can possibly be explained by domain knowledge about the book market. If I am looking for a specific book I will probably search for it by title or author. Take away these specific searches that would otherwise have used the singular keyword and you are left with predominantly plural keyword searches.  The single exception to this was searches for  ‘phone book’ which had 125 times as many searches as ‘phone books’ but is as much a special case as ’sock shop’ was above. So, knowledge of the domain that searchers are seeking to explore can, in some specific circumstances (sock(s) and book(s), for example) reveal whether singular- or plural-versions of keywords are likely to predominate.

Linguistic factors are the second way of predicting searcher’s preferences for singular or plural keywords. A keyword can be used in two distinct ways for search: it can be the primary search object or it can be the qualifier for a different primary search object. So, for example, if I search for ‘pink laptop’ the thing I am looking for is a laptop and the qualifier I am applying to my search is that I would like the colour of my laptop to be pink. If, on the other hand, I am looking for ‘laptop review’ then the primary object of my search is to find reviews and laptop becomes the qualifier - I want the reviews to be about laptops. Linguistically, the word ‘laptop’ is a noun adjunct within the phrase ‘laptop review’. When keywords are used as qualifiers (noun adjuncts) for a different primary search object, they are nearly always singular.

Keywords are singular when used as a qualifier

A related but different linguistic factor is where the construction of the keyphrase demands the plural version of a keyword to make sense. Searches for ‘lcd tv(s)’ provide a striking example of this.

Search volumes for particular keyphrases containing 'lct tv(s)'

For reasons we will explore within psycho-linguistic factors below, most searches for lcd tv(s) use the singular-version of the keyword. The two exceptions are ‘compare lcd tv(s)’ and ‘plasma and lcd tv(s)’, where plural-versions of the keywords predominate (for ‘plasma and lcd tv(s)’ over 99% of searches were plural). The reasons would seem to be that the keyphrase demands the plural version to make sense: if the phrase was ‘compare lcd tv’ it would appear to beg the question ‘compare with what?’ Similarly, searching for ‘plasma and lcd tv’ suggests an interest in a television that is both a plasma and an lcd device. Searching for ‘plasma and lcd tvs’ indicates and interest in both types of television. Finding keyphrases, such as these, where grammar requires a plural keyword within a category of search terms that are otherwise singular could provide a useful niche for search engine optimisation efforts.

The linguistic factors described above are concerned with the grammar or syntax of language.  Psycho-liguistic factors, by contrast are concerned with people’s beliefs, expectations or intentions in their use of language. One such factor is whether it is the searcher’s intention to find a single product or a range of products to choose from.

Search intentions affect use of singular or plural keywords

This data shows that searching for ‘cheapest’ or ‘best’ - where the intention is presumably to find a single product - leads to predominantly singular keyword searches. By contrast, searches for ‘cheap’ - where the intention is probably to find a range of products - leads to predominantly plural keyword searches.

Another psycho-linguistic factor appears as though it might resolve one of the mysteries of this research. The data on ‘lcd tv(s)’ stood out for predominantly using singular keywords: 88% of the total search volume reported by Google Keyword Tool was for singular keywords. Also, lcd tv(s) defied the cheapest/best=singular, cheap=plural rule just described above - searches for ‘cheap lcd tv’ outnumbered ‘cheap lcd tvs’ by almost 3:1. So what’s going on here? One possibility is that it is something to do with the particular product category - when people search for televisions they, for some unknown reason, search using singular keywords. If this was the case, the same preference for singular keywords should be found for the keyword ‘television’ (instead of ‘tv’). But it isn’t! In fact it is the opposite - 86% of the search volume reported by Google’s Keyword tool for television(s) used plural keywords. This suggests, therefore that searchers have some form of psycho-linguistic bias against using the plural version of abbreviations such as ‘tv’.

The next post will estimate the search engine optimisation opportunities presented by singular and plural keywords.

This is the first of 3 posts on the use of singular and plural keywords for search marketing. This post presents data on how searches for singular and plural keywords changes the top ten search results on Google. The second post explores search volumes for single and plural keywords. The third post estimates the potential value of optimising for both singular and plural keywords.

Knowing how to manage single and plural versions of keywords (e.g. bed and beds) is a significant issue for search marketing, if for no other reason that it potentially doubles the number of keywords being managed! It is, of course, possible that it doesn’t matter. Search engines are easily capable of returning search results for both the singular and plural keywords, regardless of which was used as the actual search term (see Google now uses stemming). The research reported below sets out to estimate how much the use of singular or plural keywords changes the search results returned by Google.

Research method

The singular and plural versions of 14 keywords were entered into Google.co.uk and the returned search results were analysed. The keywords selected were: bed(s), cabin bed(s), book(s), comic book(s), digital camera(s), best digital camera(s), laptop(s), acer laptop(s), lcd tv(s), plasma and lcd tv(s), mobile phone(s), business mobile phone(s), sock(s) and black sock(s) - to try to cover a number of shopping categories and both high and low search volumes. The searches were conducted using a Google Chrome browser in incognito mode. Following Robin Goad’s evidence-based conclusion that ‘plural terms are better at sending traffic to retailers than singular terms’ the plural version of keywords was used as the benchmark and the singular version compared to that benchmark. So, for example, a search was conducted for ‘beds’ and the domains in the top ten search results were recorded (e.g. www.bedroomworld.co.uk is in position 1). Then ‘bed’ was searched for and the first position of the domains found for ‘beds’ was located in the search results. If the domain did not appear in the first 100 results then its position was arbitrarily designated as 100 - note this underestimates how badly the worst performing sites were doing.

Results

Difference in SERP position for singular and plural keywords

This graph shows how many domains (out of 140) had exactly the same position for singular and plural keywords in the search results page (0 difference) and how many differed by 1 or 2 positions, 3 to 5 positions, 6 to 20 positions or more than 20 positions.  Whilst 44% of domains ranked equally well for singular and plural keywords (difference of 2 or less), 56% of domains ranked differently (difference of 3 or more).  59 out of the 140 (42%) domains that appeared in the top 10 search results for plural keywords did not appear in the top 10 for singular keywords.  20 out of the 140 (14%) domains appeared at least 20 places lower for singular keywords compared to plural keywords.

So, whilst search engines have the capability to return the same search results for singular and plural keywords, they have clearly chosen not to - a few domains that appear in the top 10 search results for plural keywords do not appear in the top 100 search results for singular keywords. The majority of domains (56%) rank differently (3 or more search results positions) for singular and plural keywords.

In the next post I will present results on search volumes for singular and plural keywords.

The entries on a search results page comprise a title, snippet and URL:

Search engine result for 'bbc' from Google

There has been lots of discussion within the SEO community about optimising use of the title tag, how to improve snippets using the meta description tag and, recently, about changes to Google’s presentation of the snippet in search results. What I cannot find discussed, however, is the effect of search keywords in the title and snippet on the click-through behaviour of searchers.

I collected data on 80 keywords that resulted in a total of over 14,000 visits from Google.co.uk search results pages to 3 web sites. The keywords ranked from position 1 to greater than position 20 on the results page(s) and the web sites were all successful retail sites selling goods direct to consumers. For each keyword, I recorded the number of visits attributed to that keyword from Google that was ‘non-paid’ during one week in April 2009 (from Google Analytics). I also recorded the search volume for each keyword reported in Google’s External Keyword Tool (set for the UK and reporting search volumes for March 2009). Dividing visits by estimated search volume gave me an estimated click-through rate for each keyword.  I then gave a simple score for the presence of keywords in the title and snippet of each search result (for the 80 keywords going to one of the 3 sites).

The first result shows that the presence of keywords in the title of the search result has no significant influence on the click-through rate of searchers (all differences p>0.05).

Effect of keyword in title on click-through rate from SERPs

By contrast, either partially or fully matching the searched keyword in the snippet significantly increases the click-through rate from the search result (p<0.01 for No vs Partial and No vs Yes).

Effect of keyword in snippet on click-through rate from SERPs

It is important to note that the y-axis on these graphs is a logarithmic scale - a log transformation was applied to normalise the distribution of data so that parametric tests would be valid. The increase in click-through between no keyword in the snippet and either partial or full keyword in the snippet is 550%! The 95th percentile confidence limits suggest that having the search keyword in the snippet increases click-through rate by between 268% and 1708%.

Making sense of these results

There are good reasons to expect that the presence of a searched keyword within a search result will increase click-through rate. Inattentional blindness is well established in perceptual psychology - people simply don’t see what they are not looking for. The most famous experimental proof of this is the ‘Gorillas in our Midst video‘ by Simons & Chabris. The video shows two teams of players (in black and white t-shirts) passing basketballs (one for each team) to one another. Experimental subjects were asked to watch the video and count the number of times the white team passed the ball. Whilst all subjects were able to do so successfully, surprisingly few of them noticed the person in a gorilla suit that walked into the middle of the screen beat their chest and walked off-screen. So, if I’m searching for a particular keyword, I might literally not see the results that do not have my keyword in it.   Evidence of the impact of inattentional blindness on online behaviour was demonstrated several years ago by Jared Spool of Interface Engineering Design. He gave people a goal to achieve in navigating through a web site. Their success in doing so correlated with the the presence on the home page of the trigger words they had formulated in their minds for achieving that goal.  So, the effect of keywords within search results increasing click-through rate makes sense.  What, at first glance, is more counter-intuitive is that keywords in the title have no effect but keywords in the snippet have a large effect.

The fact that the title is in a bigger font and positioned on top of the snippet might lead us to expect the opposite effect: that keywords in the title should have a bigger effect on click-throughs than keywords in the snippet. If the keyword was always found in the title, as might be expected because of the importance of title keywords for SEO, then keywords in the snippet would be the only differentiator of relevance for searchers. But this was not the case - in the sample of 80 keywords studied 20 had no keyword in the title and 17 had no keyword in the snippet. There would appear to be two possible explanations for the importance attributed to snippets by searchers: firstly, the greater amount of text presented in a snippet enables searchers to get a better understanding of the keyword in context and this, in turn provides a better estimate of the relevance of the search result. The second possible explanation relates to the fact that Google always uses the title of the indexed web page as the search result title, whereas the snippet is compiled in different ways to try to maximise its relevance - maybe searchers have learned (almost certainly sub-consciously or tacitly) that snippets are better indicators of relevance than titles.

Whatever the explanation, the evidence presented here suggests that SEO professionals should make every effort to get keywords into the snippets of their search results.