The Google Sandbox Effect has been discussed at length in our case study of a new website first crawled in May by Googlebot. We can now further the case study with indexing comparisons and discuss interesting Googlebot crawler behavior after release, at the 75 day mark, of the study website from that very confining Sandbox.
This case study is not for the faint of heart – those just launching a new web business on a new domain name with hopes of instant indexing and immediate traffic may find their website very lonely for two and a half months – if it is in a competitive market segment. You may as well plan to stay in the Google Sandbox for at least 45 days on average. If some early release stories are to be believed, search phrases nobody wants to play with are taken pity on by Google and sent home for early release.
Those non-competitive or obscure search phrases seem to be seen as good, quiet little children, playing by themselves in Sandbox playground and are sent home early on good behavior. Googlebot probably sees good behavior as playing well with others, like a good little baby domain and NOT being competitive as some young domains can be. Throwing sand in other childrens’ faces and insisting on having your site indexed, throwing sand out of the Sandbox with your bright plastic toy shovel and bucket.
Now that the site discussed in this study is out of the Sandbox, it still lingers on the playground, unable to escape the community park and leave for the business world to play with the big boys in the outside world. It does indeed take time to grow up and be the model citizen in this new search playground. Though on the first full day after this first week of being released from the sandbox, the site has gotten 68 visitors referred by searches done at Google, the first referred search traffic coming into the site. MSN has sent 8 visitors, Yahoo has sent 6, 4 came from AOL searches, 2 from Netscape and 1 from Dogpile.
The indexing behavior of Yahoo and MSN has been nothing short of bizarre with numbers of indexed pages increasing rapidly over the first two months to reflect 6,941 pages indexed until 8 weeks into this study and we outlined previously how numbers changed as you click through results pages first upward, then downward to about half the total of highest numbers listed along the top of the results pages.
It appears that Yahoo and MSN are playing on the ‘slippery slide’ in this playground, climbing to the top of the ladder of results at about 10 week mark showing 8,210 and 6,941 pages respectively indexed, then sliding down again to 3,510 for Yahoo and 373 for MSN, as of this writing two weeks later on August 6. Still, Yahoo will show you only 1,000 (100 pages) of those results and MSN will show you only 250 results, or 25 pages, no matter how many they claim to index. MSNbot is crawling the site faster and more consistently than any of the engines, yet shows by far fewer pages indexed than the others.
One of the interesting comparisons between Google and MSN in our Sandbox study is that Google will show you most of what they claim to have indexed after you click that link at the bottom of the first page showing only 3 or 4 results when you use the “site:Publish101.com” query operator then go to the bottom of the page and click the link under the line reading, “In order to show you the most relevant results, we have omitted some entries very similar to the 3 already displayed. If you like, you can repeat the search with the omitted results included.”
Go ahead and click that link, then you’ll be presented with the claimed total of indexed pages. That number has very steadily increased since Sandbox release after 75 days from first crawling of this Sandbox study site. The timing and numbers of indexed pages at Google goes upward, and ONLY upward with VERY distinct patterns noted from raw log files. Crawling schedules seem to have been established for this site by Google and indexing changes occur on a very regular schedule.
The first observation of Sandbox release was at noon on Thursday July 28, seventy-five days from first crawling by Googlebot when a search turned up 379 pages indexed with a “site:Publish101.com” query. That number increased later the same evening to 3,660 pages at a search done around the dinner hour Pacific time. Oddly, the next day, Friday July 29, the number took a slight hop upward to 3,700 pages and on the following Monday, showed 3,770 pages indexed.
That schedule and pattern have repeated on the second week of Sandbox release when a “site:Publish101.com” query produced 5,660 results from from Google for the site on Thursday August 4 at just after noon and then nearly doubled at around the dinner hour to 10,700 pages on that same query. A final check just now on Saturday shows it at 12,100 pages indexed by Google. It should be pointed out to those who wonder about the total number of pages that this is a dynamic site with a very large archive of articles that increases daily as new submissions are contributed by member authors at the site.
Those articles are added through a content management system on a daily basis by an editor who reviews submissions and processes them for approvals or rejections. Those approved are made live from the home page nightly. We’ve started doing this on the crawler’s schedules as we’ve noted very regular visits by Yahoo’s Slurp crawler to the site home page just once daily at around 5pm each evening and Googlebot visiting the home page only once, at near 11pm nightly, so we’ve instituted a midnight activation of each day’s new article submissions on the home page of the site so that none of the new pages are missed by those crawlers. MSNbot seems to hit the home page multiple times through the day, so timing is less important for MSN.
Crawler activity has been heated, with Yahoo crawling the least and the slowest, barely seeming to attempt any updates and the total of indexed pages has not changed for over three weeks since it peaked at 8,210 pages indexed and then dropped to it’s current level of 3,510. As previously stated, Slurp seems to be unhindered by any form of consistency in indexing or crawling behavior. MSNbot has crawled extensively and fairly regularly for weeks, but that odd indexing behavior is a serious flaw in their utility as a search tool.
It should be mentioned here that AskJeeves had been noted to crawl the site extensively early in this case study and displayed a very regular and consistent crawl, but stopped abruptly three weeks ago on july 13, after hitting most of the pages then available on the site. Teoma, their spider, has been absent ever since and they have not indexed this domain at all since first crawling on May 23, over 10 weeks ago. Clearly, Teoma appears to have the longest Sandbox of all the search engines.
Much has been learned in this Sandbox case study about crawler behavior, indexing delays, robots.txt requirements and index updates at each of the top three search engines. Where that knowledge leads will, of course, change as algorithms and crawling schedules are adjusted by MSN, Yahoo and Google. But valuable information has been shared that may help other webmasters to better understand each of the factors that determine the success of any website.
We’ll outline the findings in a follow-up article at the 3 month mark and explore search referrals gained as Google adds more pages and rankings fluctuations begin to level. Meanwhile, we’d like to encourage others to review traffic logs to monitor crawler behavior on other new domains to verify findings and disclose indexing behavior and timing for new domains.