GPWA Times Magaine - Issue 23 - February 2013

Now a row in this matrix will be a vector corresponding to a term, giving its relation to each document: [ ] ni i T i x x , 1, ... = t Likewise, a column in this matrix will be a vector corresponding to a document, giving its relation to each term:           = jm j j x x , ,1  d Now the dot product p T i tt between two term vectors gives the correlation between the terms over the documents. The matrix product XX T contains all these dot products. Element ( i , p ) (which is equal to element ( p , i )) contains the dot product p T i tt (= i T p tt ) . Likewise, the matrix X T X contains the dot products between all the document vectors, giving their correlation over the terms. q T j dd = j T q dd . Now assume that there exists a decomposition of X such that U and V are orthogonal matrices and Σ is a diagonal matrix. This is called a singular value decomposition (SVD): The matrix products giving us the term and document correlations then become XX T = UΣΣ T U T and X T X = VΣ T ΣV T . Let’s be honest – most of us will never run through formulas like this. If we were going to do that I have a few theoretical algorithm formulas that will make your head spin. In simple words, LSA is the process of comparing documents to find relevant keywords that normally surround the tar- geted keyword. Example: You create an article based on the keyword “lorry.” This is a word they use in the U.K. for “truck.” LSA would analyze the article and the articles/websites that are linking to it and look for related terms. If the content surrounding the target term or the links from other documents or web- sites, or both of these combined, are refer- ring to cardinals, blue jays and mocking- birds the algorithm will tag this page as the English version of a “bird” because of the related articles and LSA. If the linking documents and links reference Ford, Mercedes and Volvo, the algorithm will tag the topic as being about trucks. The reason LSA is so important to use in your content as well as in the anchor text in your back links is because of latent se- mantic indexing (LSI). These let Google figure out how to classify the document and what authority to give it, and then in- dex the page accordingly. If your entire site is about lorries you are doing well. Otherwise I would create a silo if I expected to rank well for lorries. When running campaigns that target mul- tiple countries and languages this becomes extremely important. The complexity of multiple languages, the algorithm’s consid- eration of regional and local slang termi- nology and surrounding content, and also back linksmake the taskevenmoredifficult. Now I am going to throw one more item into this confusion. Let’s say you are in Australia, you speak English and your target is the U.S. and the U.K. Trucks are referred to as “utes” in Australia. Utes are called lorries in the U.K. Lorries are called trucks in the U.S. We are trying to target truck buyers in three countries that call trucks different names. What are your options? (For all of these options you need back links from relevant sites, with relevant anchor text.) Option 1. Create an article that references each of them. Option 2. Create an individual post for each type of truck name (Truck, Lorry, Ute). Option 3. Create silos for each type of truck. Option 4. Create subdomains for each, with geographical targeting, titles and descriptions for the type of truck you are targeting. My choice is number four. As I said earlier subdomains are consid- ered stand-alone or completely separate websites and will therefore have the best chance at ranking well. They also gain au- thority points from Google. Tip 5 » Important elements in the Penguin 3.0 update* These ranking factors are tested by taking the top 50 “winners” and the 50 biggest “losers” (i.e., increases and decreases in the SERPs) in the Penguin 3.0 update. This is not definitive and it only applies to this specific update. Google has on aver- age one or two algorithm tweaks/adjust- ments every day. The significant ones like Panda, Jagger, 65-Pack and the Freshness Update are well known to most SEOs but it is impossible to know or identify them all. Every update has different targets. It is important to keep in mind that if ele- ments like social media, hosting countries and domain age are not considered impor- tant in this update, it’s because every algo- rithm targets different parts of the overall Google algorithm. These are the parts that were found to be important NOW . A. Medium importance* • Number of indexed pages – Target 1,000+ for your sites. • Deep link ratio – I keep it a 30/70 split. The 30 percent is my homepage and my primary search term. I do not use my primary search term in more than 20 percent of the anchor text pointing to my website. I do include the keyword in related keyword string links. Note – Link brand keywords to home page 40 percent max. • TitleRank and power trust – I use PRWeb and Marketwire for press re- leases that are syndicated across hun- dreds of online news sources that are considered trustworthy and authorita- tive. The title of the article and the snip- pets (H1 and H2/H3 tags) include my target keyword and they link back to a page that shares the same title, descrip- tion and links within my other pages that refer back to this page. ( Hint – Pay $299 for a Yahoo Directory Listing. You’ll never find a link from a clean PR7 website that is trusted by Google 100 percent for $25 per month.) 32 Top 12 SEO tips for 2013