google suggest scraper (php & simplexml)

Today’s goal is a basic php Google Suggest scraper because I wanted traffic data and keywords for free.

Before we start :

google scraping is bad !

Good People use the Google Adwords API : 25 cents for 1000 units, 15++ units for keyword suggestion so they pay 4 or 5 dollar for 1000 keyword suggestions (if they can find a good programmer which also costs a few dollars). Or they opt for SemRush (also my preference), KeywordSpy, Spyfu, and other services like 7Search PPC programs to get keyword and traffic data and data on their competitors but these also charge about 80 dollars per month for a limited account up to a few hundred per month for seo companies. Good people pay plenty.

We tiny grey webmice of marketing however just want a few estimates, at low or better no cost : like this :

data num queries
google suggest 57800000
google suggestion box 5390000
google suggest api 5030000
google suggestion tool 3670000
google suggest a site 72700000
google suggested users 57000000
google suggestions funny 37400000
google suggest scraper 62800
google suggestions not working 87100000
google suggested user list 254000000

Suggestion autocomplete is AJAX, it outputs XML :

< ?xml version="1.0"? >
   <toplevel>
     <CompleteSuggestion>
       <suggestion data="senior quotes"/>
       <num_queries int="30000000"/>
     </CompleteSuggestion>
     <CompleteSuggestion>
       <suggestion data="senior skip day lyrics"/>
       <num_queries int="441000"/>
     </CompleteSuggestion>
   </toplevel>

Using SimpleXML, the PHP routine is as simple as querying g00gle.c0m/complete/search?, grabbing the autocomplete xml, and extracting the attribute data :

  1.         if ($_SERVER['QUERY_STRING']=='') die('enter a query like http://host/filename.php?query');
  2.  $contentstring = @file_get_contents("http://g00gle.c0m/complete/search?output=toolbar&amp;q=".urlencode($kw));  
  3.    $content = simplexml_load_string($contentstring );
  4.  
  5.         foreach($content-&gt;CompleteSuggestion as $c) {
  6.             $term = (string) $c-&gt;suggestion-&gt;attributes()-&gt;data;
  7.             //note : traffic data is sometimes missing  
  8.             $traffic = (string) $c-&gt;num_queries-&gt;attributes()-&gt;int;
  9.             echo $term. " ".$traffic . "
  10. " ;
  11.  }

I made a quick php script that outputs the terms as a list of new queries so you can walk through the suggestions :

The source is as text file up for download overhere (rename it to suggestit.php and it should run on any server with php5.* and simplexml).

RedHat Seo : scraper auto-blogging

Just give us your endpoint and we’ll take it from there, sparky!

I was going to make one of these tools to scrape google and conjur a full blog out of nowhere, as Christmas special, RedHat Seo. The rough sketch has arrived , far from perfect, but it does produce a blog and don’t even look too shabby. I scraped a small batch of posts off of blogs, keeping the links intact and adding a tribute links. I hope they will pardon me for it.

structure

I use three main classes,

BlogMaker the application
Target the blogs you aim for
WPContent the scraped goodies

…and two support classes

SerpResult scraped urls
Custom_RPC a simple rpc-poster

Target blogs have three texts,

file contents maintenance
blog categories category you post under manual
blog tags tags you list on the blog manual
blog urls urls already used for the blog system

routine

The BlogMaker class grabs a result list (up to 1000 urls per phrase) from Google, extracts the urls and stores them in SerpResult, scrapes the urls and extracts the entry divs, stores div-entries in the WPContent class (that has some basic functions to sanitize the text), and uses the BlogTarget-definitions to post it up blogs with xml-rpc.

usage

  1.  
  2. //make main instance
  3. $Blog = new BlogMaker("keyword");
  4.  
  5. //define a target blog, you can define multiple blogs and refer with code
  6. //then add rpc-url, password and user
  7. //and for every target blog three text-files
  8.  
  9. $T=$Blog->AddTarget(
  10.  'blogcode',
  11.  'http://my.blog.com/xmlrpc.php',
  12.  'password',
  13.  'user',
  14.  'keyword.categories.txt',
  15.  'keyword.tags.txt',
  16.  'keyword.urls.txt'
  17.  );
  18.  
  19. //read the tags, cats and url text files stored on the server
  20. //all retrieved urls are tested, if the target blog already has that
  21. //scraped url, it is discarded.
  22. $T->CSV_GetTags();
  23. $T->List_GetCats();
  24. $T->ReadURL();
  25.  
  26. //grab the google result list
  27. //use params (pages, keywords) to specify search
  28. $Blog->GoogleResults();
  29.  
  30. $a=0;
  31. foreach($Blog->Results as $BlogUrl) {
  32.   $a++;
  33.   echo $BlogUrl->url;
  34. //see if the url isnt used yet
  35.  
  36.  if($T->checkURL(trim($BlogUrl->url))!=true) {
  37.    echo '…checking ';
  38.    flush();
  39. //if not used, get the source
  40.    $BlogUrl->scrape();
  41. //check for divs marked "entry", if they arent there, check "post"
  42. //some blogs use other indications for the content
  43. //but entry and post cover 40%
  44.  
  45.    $entries = $BlogUrl->get_entries();
  46.    if(count($entries)&lt;1) {
  47.     echo 'no entries…';
  48.     flush();
  49.     $entries = $BlogUrl->get_posts();
  50.      if(count($entries)&lt;1) {
  51.       echo 'no posts either…';
  52. //if no entry-post div, mark url as done
  53.  
  54.       $T->RegisterURL($BlogUrl->url);
  55.      }
  56.    }
  57.  
  58.    $ct=0;
  59.    foreach($BlogUrl->WpContentPieces as $WpContent) {
  60. //in the get_entries/get_post function the fragments are stored
  61. //as wpcontent
  62.     $ct++;
  63.  
  64.     if($WpContent->judge(2000, 200, 5)) {
  65.      $WpContent->tribute();  //add tribute link
  66.      $T->settags($WpContent->divcontent); //add tags
  67.      $T->postCustomRPC($WpContent->title, $WpContent->divcontent, 1); //1=publish, 0=draft
  68.      $T->RegisterURL($WpContent->url);  //register use of url
  69. usleep(20000000);  //20 seconds break, for sitemapping
  70.     }
  71.    }
  72.   }
  73.  }

notes

  • xml-rpc needs to be activated explicitly on the wordpress dashboard under settings/writing.
  • categories must be present in the blog
  • url file must be writeable by the server (777)

It seems wordpress builds the sitemap as background process, the standard google xml sitemap plugin wil attempt to build in the cache (takes anywhere between 2 and 10 seconds), and apart from building a sitemap the posts also get pinged around. Giving the install 10 to 20 seconds between posts allows for all the hooked in functions to be completed.

period

That’s about all,
consider it gpl, I added some comments in the source but I will not develop this any further. A mysql backed blogfarm tool (euphemistically called ‘publishing tool’) is more interesting, besides, I am off to the wharves to do some painting.

if you use it, send some feedback,
merry christmas dogheads

some thoughts on search engine marketing

I want a quick way to check the competitors in a niche, and estimate what can I spend on a search engine marketing campaign to make money in the segment, using the search engines (read: Google).

For a quick test I dive into the money segment, got the top 70 search phrases of October off of 7Search and for each get the top 100 results, resulting in a mysql database with 7.000 url’s on 3500 domains.

Then I try three cross-sections to extract the top 25 domains in the segment :

  • (100 – position) / 100 per result
  • (100 – position) / 100 per result, +1 for first 10, +1 for first 6, +1 for first three
  • AOL experience based percentage of search volume

rating by rank

If I rate the sites’ urls based on (100 – position) / 100 per result.

place 1 = 0.99, place 100 = 0, that yields this table :

domain score
www.amazon.com 51.05
www.youtube.com 34.82
en.wikipedia.org 34.04
wordpress.com 30.93
www.streetdirectory.com 29.62
money.cnn.com 24.5
www.wikihow.com 22.84
www.experienced-people.co.uk 22.29
ezinearticles.com 22.26
books.google.com 19.22
www.problogger.net 17.52
entrepreneurs.about.com 16.53
www.43things.com 14.75
moneycentral.msn.com 14.74
answers.yahoo.com 14.41

That makes Amazon top dog for all phrases containing ‘money’. I know the first pages get most traffic, and that first method doesn’t express that, so I change the routine and do my old trick.

rating by rank, bonus for front page

(100 – pos) / 100, and
for place 1-3 3 points
for place 4-6 2 points
for place 7-10 1 points

That gives this table :

domain score
en.wikipedia.org 91.04
www.wikihow.com 67.84
www.amazon.com 65.05
www.youtube.com 56.82
money.cnn.com 50.5
entrepreneurs.about.com 45.53
ezinearticles.com 45.26
www.experienced-people.co.uk 42.29
www.freemoneyfinance.com 31.9
www.streetdirectory.com 31.62
wordpress.com 30.93
www.43things.com 30.75
abcnews.go.com 30.15
moneycentral.msn.com 29.74
moneymakerinfo.blogspot.com 29.55
www.thisismoney.co.uk 27.06
www.moneymakingmommy.com

That still does not show what domains actually get the traffic I need for conversion.

It is easy to score in a low traffic niche and anyone can be a winner on long tails, “alabama ski resort”, but long tails only get you a few hits (and wonder if they are serious?). And these results doesn’t give me a clue what I can spend (or what my competition would be willing to spend) on the actual traffic spots.

So I am going to estimate what traffic every site gets. I need search volumes and percentages for the serp ranks. I grabbed the search volumes per phrase from 7Search. For percentages, Aaron Wall quotes an old AOL source on the average click through rate of search engine result pages per spot on the first page :

Overall Percent of Clicks

Relative Click Volume

  1. 42.13%, 2,075,765 clicks
  2. 11.90%, 586,100 clicks
  3. 8.50%, 418,643 clicks
  4. 6.06%, 298,532 clicks
  5. 4.92%, 242,169 clicks
  6. 4.05%, 199,541 clicks
  7. 3.41%, 168,080 clicks
  8. 3.01%, 148,489 clicks
  9. 2.85%, 140,356 clicks
  10. 2.99%, 147,551 clicks
  1. 3.5x less
  2. 4.9x less
  3. 6.9x less
  4. 8.5x less
  5. 10.4x less
  6. 12.3x less
  7. 14.0x less
  8. 14.8x less
  9. 14.1x less

1st page totals: 89.82%, 4,425,226 clicks
2nd page totals: 10.18%, 501,397 clicks

That was what i was looking for. Given these percentages I can estimate the traffic any spot on the search engine result front page in Google would generate, and that yields a more realistic table of the money segment :

domain traffic result result+(1,2,3)
www.bidvertiser.com 737478 6.65 17.65
money.cnn.com 294316 24.5 50.5
en.wikipedia.org 160370 34.04 91.04
www.wikihow.com 132332 22.84 67.84
www.wealthsuccess.usana.com 123503 9.4 24.4
entrepreneurs.about.com 117069 16.53 45.53
www.moneymakingmommy.com 112507 11.65 26.65
www.freemoneyfinance.com 102122 12.9 31.9
www.experienced-people.co.uk 86498 22.29 42.29
www.forbes.com 84393 7.22 19.22
www.netjobs4all.com 75868 4.61 12.61
makemoneyforbeginners.blogspot.com 75399 6.53 15.53
www.missingmoney.com 65326 3.63 12.63
www.youtube.com 60586 34.82 56.82
technology.timesonline.co.uk 55009 3.53 8.53
moneymakerinfo.blogspot.com 54431 13.55 29.55
www.moneyclaim.gov.uk 50544 1.9 5.9
moneycentral.msn.com 39399 14.74 29.74

That shows which sites actively target and get the traffic in the segment, a site like amazon doesn’t show in the top of the last table.

One way of testing the validity of the estimates is comparing with alexa rank, doing a quick SeoQuake toolbar check :

domain est. search traffic alexa rank
www.wealthsuccess.usana.com 200000 20000
moneymakingmommy 110000 70000
freemoneyfinance 100000 60000
makemoneyforbeginners.blogspot.com 75000 140000
problogger.com 6500 43000

That’s roughly correct based on old AOL percentages (well done, Aaron) and Oktober’s search volumes. It does not work for youtube, msn and others that have an alexa rank based on the entire domain. Being a search engine traffic estimate it doesn’t cover your ‘audience’ (returning visitors on bookmarks and referrals off of other sites (which are included in the alexa ranking) or other segments.

Problogger for example, according to that calculation would get about 6K hits through Google but the Alexa rank indicates it has ~~160K hits a month. That indicates high return visits and direct traffic, a high usability and I think they get traffic from other segments (‘blog’ ?).

The use of it

Why did I start doing this ? To estimate a budget for niche penetration.

If I were interested in a niche and have a 1% conversion at $100,- per year per conversion, what would I be able to spend to get Bidvertisers’ traffic in the search engines ? With their traffic and my conversion, 730.000/100 = 7300 conversions per month =y 88.000 a year @100/conversion = 8.8 million dollar gross revenue on search engine traffic conversions.

If my marketing budget is 20% of gross sales, for a year project that amounts to 1.7 million dollar. If I already had 300K hits my budget would be a million.

That (roughly) answers my question.

I have my doubts about the use of a tool like this, it could function as quick way to scan a segment, make a ‘heatmap’ and pinpoint the soft spots that are easy to penetrate and get a foothold, but that would require some stiff programming and a lot of switches, as well as processing a lot more segments. Nice project for the winter.