Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.
I wish diagnosing SEO   issues was easy.
But it’snevereasy...
We need more data, less opinion.
Someone told him  this was a good        idea.
data
How to debug SEO
I even went to the length ofunderstanding the process ofelimination by building my ownphp app.
Caveat : You need to have acertain level of knowledge.
Randsaid thisin2011, and it wasalreadya big list
The debugging process.The following is based on real life events.          I’ve always wanted to say that.
Here’s the overall organic graph for2012 versus 2011. Seasonalfluctuation seemed unlikely.                             201...
If this was programming, we’d get an error   message telling us where the problem is.      Looking for problems in SEO    ...
You can use the Content > ContentDrill down report, but it only showspageviews, not visits.
If you’re usingGA, set up acustom reportusing thesesettings.
/product/                          When folders are not                          constant /city/,                         ...
It turns out that it was the/region/ city pages.
These aretopconvertingrevenuepages!
Step 1:Explain the problem clearly   with specific details.
/product/region/city/dropped 50% organic visits comparing from June to          May.
Step 2:Gather data and investigate.
Data Gathered                         Tool(s) used                     Double check with:Before / After traffic metrics (v...
The process of elimination, thinkof it as Logger.log();
Navigation         Same            Navigation    Heading 10%        Different       Heading 10%                        Sam...
Now I know what the problem is, and I have data to explain.http://www.diffnow.com/
These pages were 90%identical to each other AND  Panda 3.8 & 3.9 updates rolled out around the time           of loss.
Step 3Form a hypothesis that might    explain the problem
If we make these pages at    least 50% unique, they     won’t be considered  duplicate and will regaintraffic after a Pand...
Step 4Conduct an experiment to test      the hypothesis.
Withoutproof..
Don’t expectanyone to dothe work.
We created content to makea small sample set of pages     50% unique (5k).
We conducted the experiment and wenow have proof (data) that it worked
Step 5:Based on your experimentresults, revise or validate       hypothesis.
It’s ok tofail – buttry tominimizethedamage.
We can make SEO aMy client
Keep testing, keep            breaking stuff, be aMy client            GREAT SEO
DAVID SOTTIMANO Lead Consultant, Distilled David.sottimano@distilled.net @dsottimano
Image credits   http://eastcountyoralsurgery.net/wp-content/uploads/2011/06/kid-diver.jpg                        amateurex...
Think like a developer  debugging seo - be wizard 2013 rimini
Think like a developer  debugging seo - be wizard 2013 rimini
Upcoming SlideShare
Loading in …5
×

Think like a developer debugging seo - be wizard 2013 rimini

  • Login to see the comments

Think like a developer debugging seo - be wizard 2013 rimini

  1. 1. I wish diagnosing SEO issues was easy.
  2. 2. But it’snevereasy...
  3. 3. We need more data, less opinion.
  4. 4. Someone told him this was a good idea.
  5. 5. data
  6. 6. How to debug SEO
  7. 7. I even went to the length ofunderstanding the process ofelimination by building my ownphp app.
  8. 8. Caveat : You need to have acertain level of knowledge.
  9. 9. Randsaid thisin2011, and it wasalreadya big list
  10. 10. The debugging process.The following is based on real life events. I’ve always wanted to say that.
  11. 11. Here’s the overall organic graph for2012 versus 2011. Seasonalfluctuation seemed unlikely. 2012 2011
  12. 12. If this was programming, we’d get an error message telling us where the problem is. Looking for problems in SEO isn’t as straightforward.
  13. 13. You can use the Content > ContentDrill down report, but it only showspageviews, not visits.
  14. 14. If you’re usingGA, set up acustom reportusing thesesettings.
  15. 15. /product/ When folders are not constant /city/, segment by number/product/region/ of slashes (/) using regex to see the full /product/region/city/ picture.
  16. 16. It turns out that it was the/region/ city pages.
  17. 17. These aretopconvertingrevenuepages!
  18. 18. Step 1:Explain the problem clearly with specific details.
  19. 19. /product/region/city/dropped 50% organic visits comparing from June to May.
  20. 20. Step 2:Gather data and investigate.
  21. 21. Data Gathered Tool(s) used Double check with:Before / After traffic metrics (visits, Google Analytics Server logs imported into Splunk.bounce rate etc..)Response codes, link architecture, Screaming Frog, IIS crawler Chrome inspect element, manualOn page elements (titles, robots inspection.directives etc..)KW rankings, Backlink analysis Authority Labs, Searchmetrics, Custom scripts, manual Google Open site explorer, Majestic checks.Indexation, PageRank Scrapebox Custom scripts, SEOstatsGooglebot activity Server logs – with Splunk Google webcache (not perfect!)Source code similarity (scrape) ImportXML + Excel Manual checks, Text comparing online (Diffnow)Environmental activity SEOmoz Algorithm updates, Webmaster forums, other SEOs, Development queue / logs, SEO articles. Holidays, world events
  22. 22. The process of elimination, thinkof it as Logger.log();
  23. 23. Navigation Same Navigation Heading 10% Different Heading 10% SameLead generation Form Lead generation Form Sales text Same Sales text Same Footer Footer
  24. 24. Now I know what the problem is, and I have data to explain.http://www.diffnow.com/
  25. 25. These pages were 90%identical to each other AND Panda 3.8 & 3.9 updates rolled out around the time of loss.
  26. 26. Step 3Form a hypothesis that might explain the problem
  27. 27. If we make these pages at least 50% unique, they won’t be considered duplicate and will regaintraffic after a Panda refresh.
  28. 28. Step 4Conduct an experiment to test the hypothesis.
  29. 29. Withoutproof..
  30. 30. Don’t expectanyone to dothe work.
  31. 31. We created content to makea small sample set of pages 50% unique (5k).
  32. 32. We conducted the experiment and wenow have proof (data) that it worked
  33. 33. Step 5:Based on your experimentresults, revise or validate hypothesis.
  34. 34. It’s ok tofail – buttry tominimizethedamage.
  35. 35. We can make SEO aMy client
  36. 36. Keep testing, keep breaking stuff, be aMy client GREAT SEO
  37. 37. DAVID SOTTIMANO Lead Consultant, Distilled David.sottimano@distilled.net @dsottimano
  38. 38. Image credits http://eastcountyoralsurgery.net/wp-content/uploads/2011/06/kid-diver.jpg amateurexpert92.deviantart.com http://www.tommyzor.comhttp://www.nouse.co.uk/2012/12/28/5-films-that-make-my-christmas/home-alone/ www.amommyismade.comhttp://www.failepicfail.com/gymnast-fail-gymnast-parallel-bars-epic-fail-229.html

×