About a month ago, Google's quality Rater Guidelines were leaked and reported by Search Engine Land among others. I am a bit slow to find the time to analyze the confidential Google document but certainly found it interesting.

Here is my take on the information and guidelines used by Google to rate your web pages.

Relevance & Utility

No surprise that Google is most interested in relevance and utility when it comes to SERP'S. Without a doubt these are the key factors that their rater's guidelines hope to achieve.

Location & Language
Web pages should, in most cases, be matched to the query location and language.
Examples:
[George Bush] English (US) Expectation would be George Bush's official government web page.
[George Bush] Chinese (Taiwan) Expectation would be information about George Bush in Chinese language.
[George Bush] English (Taiwan) Expectation would be George Bush's official government web page.

Homonym's
Does the query have multiple possible interpretations? Is their a dominant interpretation?
Examples:
[football] English (US) Game played with a brown oval ball.
[football] English (UK) Game played with a round ball referred to as soccer in US.
[windows] English (US) A desktop operating system.
[windows] English (US) Multiple openings in a wall covered with glass.

Broad or Specific Query
Identification of query as being broad or specific in nature. Best match would be a broad web page for a broad query and a specific web page for a specific query.
Examples:
[digital cameras] Best match is a page containing multiple models and manufacturers of digital cameras.
[canon SD550] Best match is a page providing specifications and pricing for this particular model and manufacturer.

SERP Competition
If a query results in a huge retrieval of information then a single page, article, or link related to the query is not a good match.

If a query results in a limited retrieval of information then less information, a page or article, can be considered a good result.

Timeliness
Would a good result for a query change from year to year, or month to month? A query for George Bush in 1994 would most likely be associated with Bush Sr while a query in 2008 would most likely be associated with George Jr.

Query Classification
Queries are categorized as :
♦ Navigational [Ebay]
♦ Informational [tsunami]
♦ Transactional [download adobe reader]
♦ Combinations Maybe Informational & Transactional

RATINGS
Web pages returned for specific queries are 'rated' as :
♦ Vital - Someone owns this query, it is navigational, and is task language & location appropriate.
♦ Useful - Comprehensive, high in quality, authoritative, neither too broad or specific for query.
♦ Relevant - Less comprehensive or authoritative. There is another result that is 'useful'.
♦ Not Relevant - Not helpful but marginally connected with the topic. Too broad or specific, less authoritative.
♦ Off Topic - No relevance to the query.

Pages can't be evaluated if:
♦ They don't load
♦ Page language does not match query language
♦ Rater is too confused to assign a rating

Ratings are lowered in most cases for mismatched language and location. There are exceptions such as:
[queen of england] English (US) The British home page for the Queen would not receive a rating reduction.

If unsure of rating the rater is instructed to select the lower rating. So, if stuck between useful and relevant they are instructed to select the relevant (lower) rating.

Quality Score
The following quality scores are used:
♦ Not Spam
♦ Maybe Spam
♦ Spam

The quality score does not impact the rating so a Vital rating can be accompanied by a Spam quality score.

Spam scores are obtained by:
♦ Scraped, duplicate, or nonsense content with ppc ads.
♦ Parked domains
♦ Thin Affiliates
♦ Hidden text or links
♦ Javascript redirects
♦ Keyword stuffing
♦ 100% framed pages
♦ Sneaky redirects

Flags
Pages are flagged if they are:
♦ Pornography
♦ Malicious

Assuming you don't spam, serve up porno, or perform malicious acts (close browser, virus, spyware) then the important thing that can be drawn from this rater guide is to match your content with the queries that you anticipate will draw traffic to your page.

♦ Should my page actually become many pages since the SERP's are huge.
♦ Is the query broad? Make the page content broad.
♦ Is the query specific? Include specific details, pricing, specifications, reviews, anything a shopper would want to know about the specific query.
♦ Is my query a homonym? Stay away from keywords and phrases that have multiple common meanings. These cause trigger finger back button type traffic anyway.
♦ Once your keyword focused page is built, look at your log files to see if you have actually built broad pages that are somehow attracting more specific queries. Grow your site by adding sub-pages to the existing page with more specific targeted content.