際際滷

際際滷Share a Scribd company logo

Using Scientific Method in SEO

Download as PPTX, PDF
Branko Rihtman
Branko Rihtman

This document discusses Branko Rihtman and his background in SEO, science, and warzone experience. It then goes on to explain the scientific method and how to properly conduct SEO experiments and analyze data according to that method. Key steps discussed include defining a question, gathering resources, forming a hypothesis, performing experiments, analyzing and interpreting data, publishing results, and getting other perspectives on findings. Problems with SEO experimentation like lack of control and limited signals are addressed.

1 of 13
Download to read offline
Branko Rihtman, June 2010
@neyne About me Branko Rihtman SEO Scientist Warzone Experience Becks http://www.seo-scientist.com http://www.whiteweb.com
@neyne The Scientific Method 1.Define the question 2.Gather information and resources (observe) 3.Form hypothesis 4.Perform experiment and collect data 5.Analyze and interpret data 6.Publish results 7.Retest (frequently done by other scientists)
@neyne Define the question Usually done after observation of a phenomena Sudden change in locations (over-optimization penalty, brand algo, etc.) Anomalies in SERPs Someones blog post (footer links, second link doesnt count, etc.) Social media Things to remember: You dont have time to test everything You dont have the ability to test everything What is the ROI of your test ? Every change you do on your website is a potential experiment! "The most exciting phrase to hear in science -the one that heralds new discoveries- is not "Eureka!" but "That's funny...". (I. Asimov)
@neyne Gather information and resources Related blog posts Problem: non-standardized terminology -950 Penalty Over-optimization penalty OOP Unbalanced link equation Footer link abuse Sidebar link abuse Boilerplate linking Forums Problem: misinterpretation Social Media
@neyne Perform experiment and collect data Things to remember: Monitor your testing grounds Testing keyword choice: [sdkfsdjwekrwle] 0 results [payday loans] 5.5M results [translational remedy] 340K results Document everything! Problems: Lack of control over the system Limited output signal (ranking, title, snippet) Non-conformity of results No reliable controls
@neyne Solution Multi-directional experiments change - Reaction X State A State B Reaction X - change
@neyne Interpret data and draw conclusions Does the conclusion agree with expectations ? Does the conclusion have an alternative explanation ? Does the conclusion agree with other existing data ? (if not, what was different ?) Bounce the findings off someone Important: No definite conclusions!!! Only an increasing level of certainty that your conclusion supports (or defies) the hypothesis.
@neyne Data Analysis Statistical analysis Sample size Significance (one and two-tailed tests) Standard Deviation Correlation Average is not an end-all measure Get help! Talk to a professional! People will call you out on crap! (especially if it is self-serving) .
@neyne Avoiding the personal bias Jean Louis Th辿odore G辿ricault The Epsom Derby We observe what we expect to observe, until shown otherwise Ludwick Fleck
@neyne Go Social! Define the question Gather information and resources Forum Twitter Social Media Publish results Interpret data and draw conclusions Find an independent blogger you can trust with data Bounce the findings off someone Market your findings who will not steal them Market your findings
@neyne The SEO Testing Secret Ingredient Ethanol
@neyne Thank You!

More Related Content

Using Scientific Method in SEO

  • 2. @neyne About me Branko Rihtman SEO Scientist Warzone Experience Becks http://www.seo-scientist.com http://www.whiteweb.com
  • 3. @neyne The Scientific Method 1.Define the question 2.Gather information and resources (observe) 3.Form hypothesis 4.Perform experiment and collect data 5.Analyze and interpret data 6.Publish results 7.Retest (frequently done by other scientists)
  • 4. @neyne Define the question Usually done after observation of a phenomena Sudden change in locations (over-optimization penalty, brand algo, etc.) Anomalies in SERPs Someones blog post (footer links, second link doesnt count, etc.) Social media Things to remember: You dont have time to test everything You dont have the ability to test everything What is the ROI of your test ? Every change you do on your website is a potential experiment! "The most exciting phrase to hear in science -the one that heralds new discoveries- is not "Eureka!" but "That's funny...". (I. Asimov)
  • 5. @neyne Gather information and resources Related blog posts Problem: non-standardized terminology -950 Penalty Over-optimization penalty OOP Unbalanced link equation Footer link abuse Sidebar link abuse Boilerplate linking Forums Problem: misinterpretation Social Media
  • 6. @neyne Perform experiment and collect data Things to remember: Monitor your testing grounds Testing keyword choice: [sdkfsdjwekrwle] 0 results [payday loans] 5.5M results [translational remedy] 340K results Document everything! Problems: Lack of control over the system Limited output signal (ranking, title, snippet) Non-conformity of results No reliable controls
  • 7. @neyne Solution Multi-directional experiments change - Reaction X State A State B Reaction X - change
  • 8. @neyne Interpret data and draw conclusions Does the conclusion agree with expectations ? Does the conclusion have an alternative explanation ? Does the conclusion agree with other existing data ? (if not, what was different ?) Bounce the findings off someone Important: No definite conclusions!!! Only an increasing level of certainty that your conclusion supports (or defies) the hypothesis.
  • 9. @neyne Data Analysis Statistical analysis Sample size Significance (one and two-tailed tests) Standard Deviation Correlation Average is not an end-all measure Get help! Talk to a professional! People will call you out on crap! (especially if it is self-serving) .
  • 10. @neyne Avoiding the personal bias Jean Louis Th辿odore G辿ricault The Epsom Derby We observe what we expect to observe, until shown otherwise Ludwick Fleck
  • 11. @neyne Go Social! Define the question Gather information and resources Forum Twitter Social Media Publish results Interpret data and draw conclusions Find an independent blogger you can trust with data Bounce the findings off someone Market your findings who will not steal them Market your findings
  • 12. @neyne The SEO Testing Secret Ingredient Ethanol
  • 13. @neyne Thank You!