Topic: Security in CDN (Contents Distribution Networks) (Network Security Concerns for web users and policy issues )
Pages: 14, Double spaced
Sources: 1
Order type: Essay
Subject: Computer Science
Style: APA
Language: English (U.S.)
Abstract
Australian online shopping sales rise fast nowadays. Customers want to buy online for convenience, save time for leisure activities and money. Australian e-commerce businesses often failed in their attempts of being represented at the top of the list of the Google search queries. Not many authors answer the question of how to increase web search’s score. This paper explores the issues and problems of Google searches optimization for Magento e-commerce. The study narrowed to Google because it is the most popular search engine in Australian market compare to other analogues such as Yahoo and Bing.
The paper reports on the recently completed research and introduces a step-by-step guide for search engine optimization. It sheds light on the various Google’s factors, which influence the query results. The article proposes a new solution for being highly ranked by Google algorithm. The paper examines the various search optimization tools available on the market are and compares them with the proposed solution.
Most of recent approaches only consider the typical factors; they do not take into account some special factors that influence the ranking results. This gap is bridged by this research. The solution proposed in this thesis helps to deal with non-transparent, hidden, external, and controlled by the Google factors in order to improve the organic ranking. The paper will be useful to theoretics and practitioners because the ability to be displayed in the top positions in the search results leads to higher sales and business prosperity.
Research Topic: 1st page on Google ranking strategy for e-Commerce in Australia: Best practices in SEO (Search Engine Optimization) for Magento
Introduction
1.1. Background
Australian online shopping sales represented $14.9 billion with an impressive growth rate of 11.3% in 2013 (NAB, 2014). Magento is the most popular self-hosted e-Commerce software with 26.1% of the market share followed by: WooCommerce (9.7%), PrestaShop (8.2%), OpenCart (8%), VirtueMart (5.8%) and Shopify (4.5%) (AheadWorks, 2014). Top 10 world brands using Magento platform are Samsung, Ford, Fox Connect, Lenovo, Olympus, Men’s Health, Vizio, Nike, Nestle Nespresso (Mage Works, 2012). There are three main search engines: Google, Yahoo and Bing that dominate the global search market. However, a study proves that Google owns 95% of the market share in Australian Search Engines. Bing has marginal 4%, and all the rest engines combined represent less than 1% (Clicky, 2015).
Ranking on the first page of Google for e-Commerce owner correlated with high sales potential. A study shows that a click through rate for the website on the 1st page of Google represents 91.5% (Traffic, 2013). Some of the background information regarding online shopping in Australia is presented in Figure 1.
Google algorithm is a scoring model based on over 200 parameters that are called “signals” or “clues”. A website’s position in Google’s search is a score. Website’s with score 1 rank highest in search results. Marketing Companies crave to predict which parameters Google will decide to prioritize in the upcoming year. Google is changing its algorithm every single day but only does a press release for major changes in the algorithm.
1.2. Issues and significance of the study
Potential customers are most likely to purchase from the website found on the first page of Google. Australian e-commerce businesses often failed in their attempts of being highly ranked by Google and represented at the top of the list of the search queries. There are many reasons for that: not easily found missing keywords on the page, web page loading speed is slow, backlinks, thin content and so on.
SEO services specialized for Magento is a very lucrative business with relatively low competition. On Amazon, one single book is available in English by Robert Kent, Magento Search Engine Optimization (January, 2014) and rated five stars. A 2nd book was published by Holder Reibold in German only in March 2015, Magento SEO komapkt, Germain Edition (March, 2015). This study contributes to the general knowledge on SEO for Magento topic.
Two existing solutions by Backlinko (Dean, 2015) and Searchmetrics have been selected. Both studies attempt to identify the most significant ranking factors for Google search results. None, of these solutions, target SEO factors for Australian specifically for e-commerce platforms. Therefore, a practical implementation may result in high cost and little return on investment for e-commerce owners in Australia.
Dean published 200 Google’s Ranking Factors, but no one attempted to apply the factors and propose a targeted solution for Magento e-commerce.
1.3 Research aims
The primary aim for businesses to optimise Magento store on Google Search is to increase the internet sales.
There are several objectives stated in this research:
• Objective 1. Identify Google’s Ranking Factors for e-commerce: How to ensure that company’s web-site and products are shown on search results pages? Google picks and chooses what to show following its algorithm. This algorithm takes weights many factors of company’s site and products to grant a particular rank on the results of a search query. Google does not disclose its secrets, even experts with extensive experience do not know everything that goes behind the scenes. However, there are some key factors that influence the search results. To find out these factors is the way to optimize the search and raise the curtain to look what goes on behind the Google Shopping scenes.
• Objective 2. Provide different solutions for improving Magento ranking that meet Google’s guidelines: Google uses over 200 ranking factors in their algorithm. Some of the existing consultants are offering either the complete list of factors for Google search optimizations or vise verse the narrowed short list of the most weighted factors. The objective is to know, what are the existing providers of such solutions?
• Objective 3. Test the solutions by applying analytical & statistical approaches on a specific e-commerce scenario: Different approaches can be applied to tests, compare the existing solutions to realize their advantages and disadvantages. Having the econometrics expertise, I applied the analytical and statistical approaches.
• Objective 4. Provide recommendations: To achieve better search rankings is desirable for every internet business because organic search traffic is a proven driver of sales growth. Providing a step by step guide for improving Google Ranking is a way to contribute to theory and practice of existing knowledge on e-commerce and particularly within an Australia context.
1.4 Research overview
This summary of the research is an overview with regards to the unit of analysis. The solutions for Magento e-commerce for the Google search Engine are the subjects or the unit of analysis in this study. This study is explorative. Exploratory research projects emphases deep study of a few solutions (two existing solutions) and compare them with a newly developed and proposed solution. It is a combination quantitative (statistical analysis on a large number of factors) and qualitative approaches (sense making exercise on selected few factors).
This study consists of the literature review part and an empirical part. The empirical part of work devoted to the development of the Model and then the solution. By the creating the Guideline the study finalized. The research conducted with respect to the ethical principles established at CSU University.
Literature Review
2.1 Change history in Google algorithm
Google constantly tunes the ranking algorithm with over 600 changes a year. Most of these are adjustments do not announced publicly. Major SEO companies and research labs are tracking major changes in search results to predict future significant factors (Haynes, 2014).
When Google releases a major change in their algorithm, they are given an exotic name:
• Caffeine (2010): Dramatic increase in speed of regular crawling and reindexing
• Panda (23 Feb 2011): “A major algorithm update hit sites hard, affecting up to 12% of search results (a number that came directly from Google). Panda seemed to crack down on thin content, content farms, sites with high ad-to-content ratios, and a number of other quality issues.”
• Penguin (24 April 2012): Targeted data processed out the main search index.
• Hummingbird (20 August 2013): Core algorithm update that prioritises good quality content that answers people’s questions rather than the number of backlinks and poor quality content with the only purpose to rank for a keyword.
• Pigeon (2014): A dramatic shift in local geographical search results. It has very important impact on local e-commerce listing on Google search
• Mobilegeddon (22 April 2015): boosts mobile friendly website. (Moz, 2015)
That shows the importance to be aware of the new releases. Black Hat SEO has proven to be ineffective and destructive in a long run. Google applies automatic and manual penalties on players that try to trick their algorithm. Black HAT SEO refers to applying techniques that breach Google’s guidelines and artificially boost a website ranking. Providing great user experience and useful content has proven to be a much better strategy in the long run.
2.2 Google Signal Families
There are over 200 signals but Google. Google has never published a complete list. They only provide hints to their users. Here are factor families identified by Brian Dean. Check the complete list of factors for each family in the Annexe.
Domain Factors
Page Level Factors
Site Level Factors
Back Link Factors
User Interaction
Special Algorithm Rules
Social Signals
Brand Signals
On-Site WebSpam Factors
Off Page Webspam Factors
2.3 Google Algorithm vs Scoring models
Google’s Algorithm is multidimensional, multilevel, multi-scenario intelligent software. Even most of Google’s employees do not fully understand the logic of the algorithm. Last year, Google invested $309.2 million in artificial intelligence start-ups. It hints that Google is trying to replicate the human brain and logic within it’s searching algorithm (Bloomberg News, Jack Clark, 2014).
On the other hand, predictive scoring models are estimation formulas determined by a statistical approach. Credit Scoring models are highly used within banking to determine whether a client qualifies or not for a credit line. Bloomberg, Moody’s, Standard & Poors rating agencies publish sovereign debt, government and corporate credit default ratings (AAA, AA, BBB, BB, B, CCC, CC).
Google ranking algorithm falls under behaviour scoring models. (Scoring models) Several SEO research agencies have attempted to determine the factors that play an important role in Google’s algorithm without achieving stable results. Estimating predictive models in constantly changing and unpredictable environment never leads to stable, long-term significant results.
SEOMETRICS published SEO RANK Correlations and Ranking Factors 2014 for Google in US (see Annexe). These are top 12 significant signals such as click-through rate, relevant terms, Google+, number of backlinks, Facebook shares, Facebook Total, Facebook comments, Pinterest, SEO visibility of back linking URL, Facebook Likes, Tweets and no follow backlinks.
Even though Google will not share a billion dollar secret, many attempts to build a relatively accurate scoring model that mimics Google’s algorithm logic. SEOMETRIC article gained a lot of popularity in the industry, but it remains questionable if their approach is effective or does it apply in the e-commerce sector in Australia. Will this model remain stable and provide positive results in Australia must be tested.
2.4 Magento Solutions to meet Google’s guidelines and improve organic ranking
2.4.1 Magento SEO Extensions
Many business owners believe that they can pay an IT developer to set up a Magento store, and it will instantly make money. In reality, this is when the journey in business success begins. A large range of SEO products for Magento is sold by many extension providers all promising improvements in Google search.
Here’s the list of most popular providers: Amasty based in Byelorussia provided exceptional customer service. Kent author of Magento SEO (2014) also covers MageWork, Rocket Web, FishPig, AddThis and CreareGroup in his study.
Despite the fact that, these solutions have a positive contribution to the ranking score they can’t solve all SEO factors or write rich content for users.
Ongoing SEO
The ranking is under a high influence of unpredictable changes due to rising competition, external factors such as updating Google algorithms and technological innovations (ipads, phone retina displays, wearable watches etc.). Thus, to develop any long-term strategy is not possible. What is needed is continuous content creation and publishing, researching, tuning, testing, conversion and performance tracking.
SEO can have a great positive impact on company’s business and must be treated as an ongoing expense or investment. SEO is a high cost of implementation, and, therefore, constant monitoring is needed to estimate the return on investment that results in sales.
Proposed Model
This study is an example of assembling the various existing approaches (Brian Dean Backlinko, 2015; SEARCHMETRICS, 2014) and using them as building blocks to create a new model.
Existing solutions
In the figure above factors from both solutions are listed. Both authors are targeting Google.com for no particular type
3.1.1 Brian Dean Backlinko solution
Dean has not specified the methodology he used to determine the results. Analytical approach is assumed.
3.1.2 SearchMetrics
However, SearchMetrics stated that the correlations were determined using Spearman correlation technique by examining 300,000 URLs in top 1-30 results.
The following 2 very interesting and alerting statements are present on the website about the methodology used. (SearchMetrics, 2015):
Statement #1: “Correlation: An analysis used to describe differences between certain properties of URLs ranking from position 1 to 30 (without implying any causal relation between property and ranking).”
Statement #2: “Average Values: An analysis used to describe the existence or extent of certain features per position on average (allowing an interpretation of possibly more relevant factors in general).”
These statements will be explained further in the next paragraph.
Spearman’s Rank-Order correlation explained
When should you use it?
Spearman correlation is a statistical test that determines existing dependence or causality between 2 variables.
“Spearman’s correlation coefficient, ( , also signified by rs) measures the strength of association between two ranked variables.” (Laerd Statistics)
The hypothesis tested:
H0: There is no association between the two variables [in the population].
So 2 outcomes are possible:
1: Accept H0 : ρ≥αthere’s no relationship between 2 variables with α = 0.05 or α = 0.1
2: Reject H0 : ρ<α There’s a relationship between 2 variables with α = 0.05 or α = 0.1 It is the statistician’s choice to set the value to α = 0.05 or 0.1. If 2 variables are determined to be dependent there’s less than a 5% or 10% chance that the strength of the relationship found (rho coefficient) happened by chance if the null hypothesis were true. Let’s reconsider the statement #1: “Correlation: An analysis used to describe differences between certain properties of URLs ranking from position 1 to 30 (without implying any causal relation between property and ranking).” The main concern is the text in the brackets added by the author. We have just defined that the main goal of computing the Spearman’s correlation is to prove an existence of “causal relationship between 2 variables”. If the author refutes the core hypothesis of this test, then all coefficients published by SearchMetrics have virtually no interpretation or meaning. Also, SearchMetrics published calculated coefficients without the probabilities (ρ) that allow determining if the hypothesis of causality between the factor and Google score is present. Without this additional data, the calculated coefficients are nothing else than an idea of a trend or simply average or mean calculation. (Statement #2 confirmed) Assumptions Assumption (1): two variables that are ordinal, interval or ratio Assumption (2): monotonic relationship between 2 variables Let’s look back at the factors selected by Search Metrics. Some of the variables selected in their list may potentially comply with Assumptions #1 and #2: Google+, Facebook Likes, speed etc. On the other hand, many other variables are clearly binary or qualitative do not meet these assumptions: Proof of Terms, length of url, image count, etc. In conclusion, the methodology clearly violates statistical fundamentals of Spearman correlation statistical assumptions and makes it unclear how the coefficients associated with these variables should be interpreted. Is it a sign that the article targets audience with no statistical background and is mainly a sales hook to gain more subscribers without providing any meaningful content? The existence of significant correlation does not imply causality. This a well-known problem studied by statisticians known as logical a fallacy. A well-known example is: there’s a significant correlation between global warming and the decrease in number of pirates during the past centuries. Therefore, global warming increases mortality of pirates. Searchmetrics, study apparently lists correlation coefficients that are not necessary significant. So interpretation of these factors may potentially result in even greater confusion and logical fallacies. Hybrid Methodology A hybrid approach has been selected that consists of applying Spearman’s correlation test on the data that meets the tests assumptions: 2 variables are ordinal and monotonic relationship is identified. Also, probabilities will be calculated and causality between variables will be interpreted on a specific case study. If the data distribution does not meet Spearman’s test criteria analytical approach will preferred based on other statistical criteria. Implementation of the Ranking Model for e-commerce in Australia (800 words) The model is tested within Australian context. 4.1 Select a keyword to rank on Google A broad keyword associated with associated product sold in Australia was selected. It will be referred to as “mykeyword” throughout this study. Here are some examples of broad keywords: laptop, tablet, smart phone, dress, sheds, socks, etc. On the other hand here are some specific keywords: Apple Ipad Air, Nike Pink shoes. Paid subscription service by SEMRUSH tool evaluates the difficulty of the keyword to 70% with search volume of 33,100/per month, average cost per click for Google ads of $US 2.46 and high competition of 91%. There are 64.8 million results on Google.com.au for this keyword. Our goal is to rank in the top 10 results out of 48.5 million what is big challenge. (SEMRUSH, 2015) The top 5 results are attributed to 4 e-commerce stores & Cosmopolitan newspaper on the 2nd position. The table below is an estimation cost of Organic (FREE traffic) in USD. The 1st position on Google search is worth $US 90,000 monthly for this keyword or equivalent of $US 1,080,000 yearly in FREE Advertising. It receives 46% of whole traffic for this keyword. Segmented data is shown the table and Figure 5 below. 4.2 Data collection Process 4.2.1 Tools used in data collection Top 50 websites listed on Google for “mykeyword” were analysed. SEOQUAKE toolbar: free add-on for browser that allows to download search results from Google and SEO parameters such as page rank, Google index, keyword density and others into a csv file SEMRUSH: offers a large number of analytical tools Keyword Density Tool (SEO BOOK, 2015) VBA scripting used to retrieve all title tags in mass. Data correction Initial test correlation estimation and statistical analysis have shown very important standard deviation noise from non-e-commerce big brands. These are big newspapers and magazines such as Cosmopolitain, Women’s Health Magasin, News.com.au, Yahoo News, Men’s Health Magasin, Wikipedia and Youtube. These types of website distinguish themselves in search results with an extreme number of backlinks, very high performing server speed and large social presence and following. It is a known fact that Google intentionally mixes different types of content in the search results. Given an example a user searches about an Ipad with an intention to buy. The search results display a diversity of media types: official Apple website, the latest news about Ipad, blog reviews, video reviews, Wikipedia article, local retailers and phone operators. The users tend to consume a variety of useful information that will eventually lead to a purchase decision. Google representatives constantly stress that the content must be useful to its consumers. Therefore, e-commerce websites do not directly compete with Youtube videos, Big News papers or Wikipedia on the search results. These sources have influence on forming users buying decision by educating them about the product and in some cases driving traffic to a particular e-commerce website via link placement. Studies suggest that Google applies special rules for Wikipedia. In 2012, the test was conducted on a selection of 1000 keywords. In 96% of the cases, Wikipedia was ranked in top 5 results on Google (Johnson, 2015). In the graphic below (see Figure 6) the boatman represents Google and the fruits are the search results he needs to take to the market (end user). Because of limited seats on the boat the man is likely to take all types berries (Youtube, Wikipedia, News, Blogs) to the market because people like different flavours, but when it comes to plain boring carrots (e-Commerce websites) he will only select the best quality ones. In conclusion, it is important to exclude other platforms from our statistical analysis. E-commerce websites are competing among themselves for better content, fastest server, meta descriptions, number of backlinks and many other factors. As a result, only 27 e-Commerce websites were selected for the study and 23 other types of platforms were excluded. Figure 6: Google Search Logic Collected factors Here’s the list of Google factors collected for the research. Google search results can be represented by a function. Y is calculated output or the position rank. Y_i=Ϝ(x_1,x_2,…x_n ) Y_i=Google Score or Rank 1 being the highest score x_j=ranking factors The study will identify the factors that have the highest negative coefficient that has the highest causality effect for the first page positioning. .3 Test model 4.3.1.1 Social signals Backlinko study has identified very strong impact from the social factors: Facebook, Twitter, Google Plus, Pinterest. Pinrerest factor was dropped out of our analysis because none of the top 5-e-commerce websites had this profile. Data clearly shows very low social signals for the top search results. Hypothesis of social media impact on Google search is rejected. Moreover, the graph clearly shows no presence of ordinal data patterns. These types of data distribution DOES NOT allow us to apply Spearman correlation method to determine the correlation factor. Also, Google representative makes a public statement that Google does not treat Facebook or Twitter any differently form other pages. Also from engineering perspective Google does not use the number of likes, shares, tweets or retweets in the algorithm (Google Webmasters, 2014). -It was funny because there was an SEO that said: “Ok, we see a lot of links on Facebook and those are the pages that rank really well. “ -But that’s correlation. That’s not causation. Instead it’s probably that there’s something, and because there’s something awesome, then it gets a lot of likes on Facebook and a lot of people decide to link to it. That’s the sort of thing where the better content you make, the more people are to like it not only in Google, but in Twitter and Facebook as well. It is clear that many SEO agencies and experts do not fully understand the difference between statistical correlation and causation factors. Having strong Facebook, Twitter and social media profiles in general are great tools to drive direct traffic and brand awareness. But the number of likes will not make a page rank on the first page of Google because these variables are independent. 4.3.1.2 Backlinks Search Metrics study determined high importance of number of backlinks alone. On the other hand, Brian Dean specifies that relevant backlinks to the industry are important. The data distribution is ranked and the shapes of the graphs are relatively monotonic. Spearman’s correlation test can be applied: The correlations and probabilities are all listed in the first column. All probabilities fall under 10%, so they are all significant at 10% error that the correlation happened by chance. Referring domain, referring ips, follow links, total backlinks and no follow links all fall under 5%. These variables are considered to have a dependency with Google_score variable with 5% error that the correlation happened by chance. Follow_Links have a probability =0.09 so it must be accepted at 10% error allowance but rejected at 5% error criteria. There’s no doubt that backlinks play an important role in Google ranking algorithm. The Internet is full of black hat and white hat SEO backlinks building services. It is very important to do the research before hiring an SEO company as many services claiming white hat technique may result in greyish outcome. Google is getting smarter at figuring out high-quality natural links from spam and unnatural link building. Also, Google masters advise marking as no-follow all links from websites where “commercial relationship” is present. (Google Webmasters) The report “10 Big Brands That were Penalized by Google” covers Interflora, flower company that was unlisted from Google for sending free bouquets to bloggers who linked back to their website (Marketing Land, 2014). There’s no doubt that link building has an impact on your ranking. But how many links do you need? 1st page Google (3300 total backlinks): Google free traffic evaluated to $90K/month or $1080K/year website. SEMRUSH allows to see all the backlink profiles. It is clear that a huge number of these links follow and no follow are spam linking from international websites from Dubai, Korea, China with irrelevant content to the industry. A number of relevant blogs with paid advertising are detected as well. 3nd page of Google (5500 total backlinks): during the backlink profile analysis of this website a threat on Google Webmaster forum was found published by the founder of the website on April 28, 2014. He claimed that his website fell from the 1 to 30th position on Google. It has been identified that his backlink profile grew very fast within a few months reaching 7000 backlinks. The owner claimed that he never hired any services to create these links. It’s clear that he recovered his position since then placing him 3rd of Google Search. He receives free traffic worth of $26.8k/month or $322k/year. The spam link structure is similar to the first website. 4st page Google (80 total backlinks): This website has exactly 98,54% fewer backlinks than it’s predecessor. This is a huge drop what makes us wonder: is it quality of quantity that really counts? This website receives free traffic worth $17.5k/month or $210k. The link analysis revealed a huge number of spam backlinks with a similar structure to the previous 2 websites. This leaves us wondering if Google really penalises websites for spam links or rewards them? Or are there other factors that are more important in ranking? Considering that all these spam links do not add any value to the users, Dean’s recommendation creating relevant high-quality links is probably a better long-term investment. The figure above does not allow to easily identifying if the speed has a positive effect on Google ranking. This data distribution can be used for Spearman rank order calculation. Desktop speed correlation coefficient=0.22 and mobile speed=0.05 are positive. Does this mean slower loading speed of e-commerce websites rank higher on Google search? The probability calculated: Google_score vs mobile speed = 0.80 >0.05 we accept H0 mobile speed does not influence google_score ranking
Google_score vs desktop speed = 0.26 >0.05 we accept H0 desktop speed does not influence google_score ranking
Mobile speed vs desktop speed=0.0005<0.05we reject H0 in favour of H1: mobiles and desktop speed are dependent variables what is logical considering both variable are influenced by on the same infrastructure such as server speed, VPS, Gzip compression, caching and other factors.
Therefore, the hypothesis that slower website will provide higher ranking on Google is false because these variables are independent.
Of course, this hypothesis is not logical. Google does not reward websites will low performance as it negatively impacts crawling speed for Google bots. On the other hand, driving a Ferrari does not guarantee you the first page of Google simply because you can afford the price.
So should the speed factor be rejected all together? Absolutely not!
Multiple studies showed every second of loading time for e-commerce counts in conversion rate per customer. Decreasing the loading speed results in increase of lost sales. “If an e-commerce is making $100,000 a day, a 1 second page delay could potentially cost you $2.5 million in lost sales every year.” (Kiss Metrics)
The Figure 15 shows Google analytics from the tested e-commerce website. Customers are very impatient while doing their purchases, so are the Google bots. Until, end of may the e-commerce was hosted on an optimized webserver. Google bots were spending little time downloading a page but in kilobytes and number of pages crawled was significantly higher. Since migration, the opposite is true. Google bots spend more time downloading a page but compensate with less kilobytes downloaded and less pages crawled. If it’s hard for bots to find something they index less.
Figure 15: Google bots crawling statistics
In the figure 16 statistics were calculated on the sample data. I would recommend to have the fastest website as possible but it generally comes with a high price tag. If the price was an issue I would recommend to achieve a at least values above the Median: 59 and 67 respectively for mobile and desktop devices.
From the observation above a noticeable pattern shows that websites with “https” visible in the url address tend to rank higher. Https is an indication of installed SSL certificate that comes at additional cost to the business owners to provide additional of security in order to protect client’s personal data. There are different methods of installing an SSL. Most popular is applying the https only on webpages on the checkout shopping cart pages where private data is entered. Loading the SSL badge on every page significantly decreases the page loading speed. There’s a clear trade off between security and speed. All top 10 websites have SSL certificates but only a few decided to make https visible on every page.
I would recommend testing “https” displayed in URL because in August 2014 an official announcement has been made that Google is using https as a ranking signal. “For now it’s only a very lightweight signal — affecting fewer than 1% of global queries, and carrying less weight than other signals such as high-quality content — while we give webmasters time to switch to HTTPS. But over time, we may decide to strengthen it, because we’d like to encourage all website owners to switch from HTTP to HTTPS to keep everyone safe on the web.” (Google Webmasters, 2014)
It is clear that this criterion does not have a strong impact on all types of websites. For example, blogs where readers do not make any purchases should not be affected. On the other hand Google clearly states that “client’s security is Google’s priority” so it stays logical that safer e-commerce websites may receive an additional boost. Due to limited time scope this method cannot be live tested.
The proportion of domain name “.com.au” largely outweigh the number other domain name extensions. Australian domain law is very strict and does not allow it’s owners to purchase domain name privacy like “.com”. It is not allowed to purchase “.com.au” for longer than 2 years. The Australian body requires a systematic check that the owner remains a legal Australian resident to continue operating the chosen domain name. This may add an additional trust signal in these domains. This remains a hypothesis and has not been officially. On the other hand, it is a proven fact that geo targeted domain names tend to rank higher on local search results.
The test above shows causality between the following variables:
Google score is influenced by the number of unique words. Correlation factor =-0.33 and the probability is 0.09. So we reject the hypothesis H0 of independence of these variables with an error of 10% that this correlation happened by chance.
Google score is influenced by the number of total words count. Correlation factor =-0.39 and the probability is 0.045. So we reject the hypothesis H0 of independence of these variables with an error of 5% that this correlation happened by chance.
Google score is influenced by the number of total words count_including_stop_words. Correlation factor =-0.32 and the probability is 0.0085. So we reject the hypothesis H0 of independence of these variables with an error of 1% that this correlation happened by chance.
This test clearly shows that longer content ranks higher on Google.
21: keyword density distribution
The top 5 results have keyword density in all text under 2.66% and in body under 3.87%. Longer content, higher occurrence of keyword and low keyword density must be perceived as richer useful content to the final user.
Content length is retained is the most important ranking factor.
4.3.1.7 URL LEVEL
From the statistical point of view 0 level url pages that are home pages tend to rank higher. It is an indication that when it comes to e-commerce Google gives preference to highly specialized shops being the best for this particular product.
It suggests that e-commerce sites that are ultra specialized have more chances to have higher ranking than those that have a “supermarket” approach. A customer looking to buy a bag of dog food is likely to receive a better experience shopping at a pet shop rather than Woolworth’s isle number 21 middle section next to the garden equipment.
This concludes to the following idea: home page is like a shopping vitrine in a brick and mortar shop. The e-commerce owner must treat the home page as their most valuable real estate. Selecting one single keyword with high ROI is a much smarter strategy than trying to rank for many keywords from a single page.
No significant patterns hints that websites with sitemap.xml may receive a boost in search ranking. This factor was not listed by Backlinko or Dean.
Casey Henry’s experiment has showed that Google indexes new pages faster on websites where sitemap.xml was installed. It only took 14 minutes vs 1375 minutes (23 hours) to reindex a new page. (NeilPatel, 2015)
There’s no proof of causality existence between the speed at which Google discovers the content and the ranking score it will attribute it on Google search. Therefore, this factor will not be treated as a priority.
All top 10 results has shown to be mobile responsive. A quick test was performed on a Galaxy Note 4 mobile device. Google clearly specifies which websites are mobile friendly directly in search results. The website on position 16 on the desktop did not have “mobile friendly” displayed and the position dropped to 22nd position losing dropping to the 3rd page of Google from 2nd page.
It comes to conclusion that since the roll out of Mobilegeddon update (22 April 2015), Google does apply a boost to mobile friendly websites.
4.3.1.10 Metakeyword tag
As of March 19 2012, Matt Cutts, Google representative confirms that the Meta keyword tags are no longer used by Google or any other search engines because of the previous keyword stuffing practices.
On the other hand meta titles and meta descriptions are very important because they appear in search results snippets and directly affect the click through rate. (Google Webmasters, 2012)
The results below prove that 43% (12 of 28) websites owners committed the mistake by filling out these keywords. Unfortunately, I’m feeling guilty by being part of them. When it comes to SEO efforts filling out fields that don’t count is an additional cost in labour and/or time. It clearly has no positive effect on SEO and probably decreases the ranking by boosting the overall on page keyword density higher than it should be.
The lesson learnt from this test, is it not to fill out the metakeywords field. This advice alone will save time and money to approximately 43% of the e-commerce owners.
. Findings and Recommendations
5.1 Recommendations
This chapter reports and grounds the findings and formulate the recommendations. The proposed work will be shortly titled as “SEOMeC” (Search Engine Optimisation for Magento eCommerce). The SEOMeC solution tested by applying analytical & statistical analysis and compared with the two solutions currently available on the market by Brian Dean Backlinko (2015) and SEARCHMETRICS (2014). The findings considered from the various points of view (platform, cost, methodology, factors, advantages and disadvantages) and reported below.
Platform: The both existing solutions (Brian Dean Backlinko and SEARCHMETRICS) are compatible with all platforms, while the SEOMeC is suitable for the eCommerce platforms only.
Cost: The SEOMeC cost of its implementation is Medium similar to solution by Brian Dean Backlinko. SearchMetrics cost takes in account many factors that do not have direct causality on Google ranking. Therefore, it’s a lot more expensive and does not guarantee fast results.
Methodology: Methodology of the solution by Brian Dean Backlinko is not defined. The deficiency of the SEARCHMETRICS methodology is that statistical correlation factor on 300,000 URL appearing in the top search for unknown keywords. The Spearman’s correlation does not include probability test and does not allow to determine the causality between the variables. The advantage of the SEOMeC’s methodology is in a combination of analytical and statistical approaches (on 1 selected keyword for top 50 search results excluding all). Only 28 eCommerce websites were included in this study. Some of the websites such as Youtube, big newspapers, wordpress blogs, aliexpress were excluded. The SEOMeC solution is currently targeted on Google.com.au search engine. Later, in the next versions it can be improved to cover whole Google.com search engine. The Spearman’s probabilities were calculated on data that complies with the assumptions and clearly determines causality between variables that positively impact Google ranking.
Factors: The specifics of the SEOMeC solution is that approved factors have the following characteristics: content length, geo targeted, do not fill out meta keywords, specialize on one keyword for the home page, speed, mobile responsive, SSL https visible in the browser, only relevant quality backlinks. The SEOMeC solution absorbs all factors available in Brian Dean Backlinko (2015) solution. Moreover, some additional factors are available in the SEOMeC solution such as HTTPS/SSL visible in URL, GEO TARGETTING, and niche specialization. Compare to the SEARCHMETRICS (2014) solution, all content oriented factors such as title Tags, relevant keywords, content length, site speed and Tweeter were taken into account in the SEOMeC solution. However, some of the factors available in SEARCHMETRICS (2014) solution were rejected and not considered in the SEOMeC solution. These factors are number of links, social signals from Facebook, Pinterest, and Google Plus.
Disadvantages: Considering limitations of the solution by Brian Dean Backlinko (2015) is that it does not target eCommerce. As for disadvantages of the SEARCHMETRICS (2014) solution, it can be stated that it over hypes the impact of social media. In addition, it does not mention mobile friendly responsive design and https as a ranking factor. Calculating Spearman’s correlations without respecting data assumptions or interpreting the causality effects introduces more misperceptions than utility to its end user. The SEOMeC solution also has some limitations, namely, some factors were excluded because it was impossible to evaluate them without ownership of the websites: dwell time, click through rate.
Advantages: The strategy proposed by Brian Dean Backlinko (2015) more content and quality oriented potentially will give great results if applied. As for the SEARCHMETRICS (2014), an advantage of covering a large list of factors could be vice versa considered as disadvantage because it is purely statistical not platform specific approach, where large standard deviation in factors may be present leading to a lot of noise in the data and making the results not ultra targeted or effective. The advantage of the SEOMeC strategy is in getting fast results at lowest cost. In addition, it is targeted for Australian eCommerce.
The recommendations for improving Google Ranking: The recommendations for improving Google Ranking based on the Case study on eCommerce within Australian Context have been formulated as the Guideline that consists of the following five steps:
1. Select a keyword “mykeyword” related to a product sold in Australia online
2. Search for “mykeyword” on Google.com.au Retrieve the data into a csv file from Google for top 50 results and identify Google ranking factors with help of software: seoquate plugin, SemRush, Keyword density tool, VBA scripts & manually retrieved.
3. Clean up the data from non-ecommerce websites (Wikipedia, Youtube, Newspapers and similar) where Google potentially applies special rules and introduces a lot of noise in data distribution.
4. Import data into eViews statistical software. Analyse data for patterns, correlations, factors significance and causality, variance etc.
5. Interpret and analyse the data based on latest facts and from statistical perspective.
6. Provide recommendations for eCommerce owners in Australia to improve the ranking.
7. Implement the solution and track the ranking changes.
5.2 Forward Testing and Findings
In financial jargon backtesting consists of applying an evaluation strategy on historical data during a specific period. This allows measuring the performance and stability of the model.
On the other hand, forwardtesting performance is using live market data ad the system’s logic. (Investopedia) Considering, Google algorithm is an unpredictable environment this method has been chosen. An e-commerce that never ranked for “mykeyword” in first 100 pages was tested for a period of 2 weeks. Due to limited resources and timeframe only a few changes were made based on the factors above:
Homepage titles modified to position “mykeyword” in the first position. The whole strategy being ranked for one single keyword with the highest return on investment.
Metakeywords field that previously was filled out was emptied.
Gzip compression activated to increase the speed. The speed remains relatively slow on the new server. Only 58/100 on desktop and 47/100 on mobile device.
Homepage lacked of content mainly images, products and a few titles were present. Due to limited timeframe, a poorly written text was added to the home page. The current word count 462 (688 including stop words) and unique words: 191 (258 including stop words). This is significantly shorter than the content listed on the first page of Google. “Mykeyword” is repeated 20 times and the total density is relatively high 5.81%.
The website was tracked using SEMRUSH after these changes were implemented. After 2 weeks, the website still does not rank in the first 100 for “mykeyword” alone. But it appeared on the 6th page of Google on position 52 for a Conclusion
The changes of a search algorithm logic managed by the Google are impossible to predict. These changes occur frequently, unexpectedly and constantly. To identify the strong ranking factors in advance means being “forewarned” and therefore “forearmed”. Goggle environment can be characterized as a highly complex, versatile, and unpredictable, where changes and their consequences could not be considered in advance. Moreover, not only Goggle environment is turbulently changing. The behaviour of modern society has been changing as well. Nowadays, consumers prefer to save time and money by purchasing goods by Internet. The highly responsive and effective solutions, which can mimic Google’s algorithm, are in demand. More merchants are needed expertise in SEO solutions for e-Commerce with the predictably high scope in the organic ranking in Goggle environment. Online shopping sales stimulate the growing demand for Magento solutions and SEO services.
There are no much existing papers which devoted to the issue investigated in this thesis. No many authors answer the question of how to increase web search’s score. This paper is aimed at bridging this gap and makes contribution to theory and practice.
Theory: I framed this thesis as a theoretic paper (reporting on a completed research) that could be useful for theorists by offering a model for solving a highly complex problem. It helps to gain new insights by developing new model and offering a fresh, simple, cost effective solution. This research helps to understand deeply the issue in Australian context through combining several existing approaches. International community also might find beneficial get acquainted with the theoretical model introduced in this report.
Practice: It is also could be useful for pragmatic practitioners. Effectively dealing within a Google search environment is far from easy and businesses often fails in their attempts to develop a guideline for enhancing the web search’s score. Ideas, insights, observations, and findings joined together, and this clear presentation as a single solution helped to simplify the complexity. A substantial contribution to practice made by developing step-by-step guideline on how to improve organic ranking. The reflection on practice and an experience in this field, helped to identify effective patterns and convert tacit knowledge into an explicit format of the guideline. The solution proposed in this thesis helps to deal with non-transparent, hidden, external, and controlled by the Google factors in order to improve the organic ranking.
Due to resource and time constraints, the scope of this research is limited. Some factors were excluded because it was impossible to evaluate them without ownership of the websites: dwell time, and click through rate. Future research needs to address these factors. Only English literature used in this research. A larger number of books and existing papers on search engine optimization might be involved in the scope of this study in the future (French, Russian, Spanish known by the author of this thesis).
Practices on Google search optimization can be influenced by local and national Australian context. The study does not include international context. That limitation provides an opportunity to repeat the study in the international context.