Earlier this month, Google invited the public to take their next generation of web search, code named Caffeine, for a test drive.

The new search infrastructure is the beginning of Google’s advance towards improving indexing speed and scale as the size of the web grows increasingly cumbersome. Google is seeking feedback on the changes from experienced users and web developers by making Caffeine available via a web developer preview.

From the official blog post:

“It’s the first step in a process that will let us push the envelope on size, indexing speed, accuracy, comprehensiveness and other dimensions. The new infrastructure sits ‘under the hood’ of Google’s search engine, which means that most users won’t notice a difference in search results. But web developers and power searchers might notice a few differences.”

What Is It?

According to Matt Cutts of Google, Caffeine is essentially a rewrite of the search index and it roughly compares with the Big Daddy index of late 2005 / 2006. In other words, it’s a BIG change to Google search.

Here’s a couple of grabs from Mike McDonald’s video interview with Matt Cutts on the subject:

“We’re shooting to get results identical to previous version. We’ll open up a few datacenters with it first and then roll it out.”

“Caffeine will be more powerful, flexible and robust – allowing Google to index faster.”

“(Caffeine) builds a powerful foundation for including any changes we want to do with indexing. Not so much for taking advantage of semantic, real-time indexing, but for getting good infrastructure in place for growth and unlock more power.”

“Webmasters shouldn’t be concerned. Caffeine does not affect your site architecture“

Something that Matt Cutts hasn’t mentioned but has been discussed amongst my colleagues is whether the Caffeine rollout is at all related to Google’s BigTable technology.

Bigtable is a distributed storage system for managing structured data, designed to scale to accommodate huge amounts of data. Google uses BigTable to store data from their various load heavy apps such as Google Earth and Google Finance. It makes sense that this would eventually roll out to search. Perhaps Caffeine is the new algorithmic skin for the BigTable search infrastructure?

Search Technology Testing

In the SEO industry, we’re so used to Google rolling out algorithm changes without fanfare and reacting to them as we realize something has shifted that this announcement came as quite a surprise to me. Paul Carpenter made the same point on the DaveNaylor.co.uk blog:

“… soliciting direct feedback from users before changes are made is something I can’t recall Google embarking on before.”

My first thought was that this was a knee-jerk reaction to the Yahoo / Bing announcement last week. But in his Caffeine blog post, Matt Cutts insists that the announcement had nothing to do with Binghoo and that they’ve had engineers working on it for months. He says that Summer is simply a good time to roll it out for testing.

So I decided to conduct my own test to see if I could notice any changes.

The Experiment

I decided to compare de-caffeinated Google against caffeinated Google using five main benchmarks:

A) Index size
B) Speed
C) Site rank
D) Link type
E) Keyword density

My tool of choice for the comparison is Facesaerch’s Caffeine Compare. The search queries I decided to test were:

1) “iPhone cases”
2) “Les Paul”
3) “diamond earrings”
4) “Kalena Jordan”

See the Detailed Search Comparison Results Chart at:

Conclusions

• Probably the biggest eyebrow raiser for me was the marked jump in keyword density between SERPS on the old Google and SERPS on Caffeine. In nearly every comparison, the Caffeine SERPS featured site titles and snippets with a much higher phrase and/or keyword density. Coincidence? I doubt it.

• It’s definitely faster. Every search query I tried on Caffeine was returned at a faster speed than with the current Google. Impressive.

• Caffeine seems slightly fresher. Some of the results I observed in Caffeine SERPS and not in regular Google SERPS were more current. For example, blog posts published within the last couple of days.

• Apart from the ego search, old Google out performed Caffeine in the index size category. But this is likely because only a handful of data-centers have Caffeine on board so far.

• Caffeine definitely has a heavier emphasis on social media, with results from sites like Blogger, LinkedIn, Facebook and Google Profiles featuring more prominently, particularly for name searches. Wiki pages still seem to rank highly in both Caffeinated and Decaffeinated Google.

Other Observations

Interestingly, a couple of other bloggers have observed different trends in Caffeine SERPS. In his blog post on the subject, Paul Carpenter says:

“… maybe blended results are getting a little less prominence. Certainly some news and image results are appearing further down the page in Caffeine than in the regular, *decaffeinated* results.”

Personally, I didn’t notice this. In fact with product related searches, I saw more blended results with Google shopping links often ranking higher in Caffeine.

But Everflux is influencing Caffeine results too, as Matthew Rogers of EndofWeb found:

“The results for any search shift and change on a daily basis, because live-search results are added to the mix, causing a more fluid day-to-day search experience along with providing more relevant data upon request.”

Comparison Tools

Want to conduct your Caffeine comparison testing? Here’s a couple of tools to use:

Facesaerch’s Caffeine Compare

Black Dog’s Compare Google Caffeine

Doubleshot’s Get Caffeinated

Search Engine Optimization is at the same time a simple concept (help the search engines find and rank your pages) and a very complex one (proper use of redirection when changing domain names, Google downranking, duplicate content and hundreds more topics that are covered online in many places and also at conferences like the SES Conference Series, Webmasterworld PubCon, or the SMX Conferences.  

Arguably the best source for basic SEO information is Matt Cutts’ blog, and he always has great summaries of the conferences at which he gives talks.    Here’s a great post from Matt today after Danny Sullivan’s SMX Seattle Conference.   Google has added some information to their famous (and infamous) webmaster Guidelines, which should be read by every webmaster as they are the best *basic* information about how to structure a site to be ranked properly.   You’ll also want to read Matt’s SEO posts which offer a lot more specifics and technical advice.  

Although several years ago you would *also* have been well advised to read up on some of the tricks of the trade such as various schemes for keyword optimization, I would argue that for most webmasters tricks are more likely to be counterproductive than productive.   This is a really rich topic because there remain many techniques that fall into a sort of gray area of optimization where ranks are affected, but crossing the Google draws between acceptable techniques and unacceptable can lead to severe penalties.   Since Google does not draw a clear objective line we have the ongoing gray area of optimization. 

Many SEO techniques relate to *linking* strategies and *keyword optimization*.     It is an area where I believe Google has in many ways fueled the rise of the very content they hate by making the rules too vague and (more importantly) allowed adsense advertising on pages that don’t meet reasonable web quality standards.   Early in the game I was often frustrated when I would improve on a bad page only to have it drop in ranks due to algorithmic quirks.   I soon decided to leave crappy but high ranked pages alone, fearing they’d be downranked if I changed them.  This in turn caused problems as Google tightened up quality standards. Google is great about transparency in several areas, but algorithmic search penalties are not one of them.

I should also say there are some exceptionally good SEO folks out there who always have amazing advice when I bump into them at conferences.    David Naylor and Aaron Wall, and Todd Malicoat all have remarkable insight into the complexities of Google ranking as does Vanessa Fox who used to work for Google and Danny Sullivan who runs the SMX series of SEO Conferences.    My general advice about SEO is to do it yourself or in-house, but there are a handful of people like this who know the game so well that the normal rules about avoiding SEO folks do not apply.

Dave Naylor (Daven) says “We all know Google doesn’t know how to count” and he’s right.

Debra Mastaler wants Google to use its official communication channel and I think most of us can identify with her.

Todd Malicoat: ”I’m proud to work with Aaron Wall and Scott Smith“. Mazal Tov, Todd. We’re proud of you too.

Jim Boykin talks about his internet marketing ninjas training program and I’m a little bit sad of the fact that everyhing turns to be commercial at the end (if not in the middle of the way). I would surely miss the free SEO tools of We Build Pages.

Rand Fishkin offers SEOmoz’s website stats for 2007 and I must say that SEO-beginners can learn alot of what interesting factors should be looked at in a website stat from scanning SEOmoz stat’s titles plus I wonder if the the list of SEOmoz’s top referring domains is the real one…

Michael Gray mentions that long URLs are problematic for social media and I want to thank him ’cause I didn’t know 225 is the limit number I should keep in mind.

Aaron Wall writes about optimizing automated communication strategies and I get to recall once again why the seo-book is one book I can read again and again and again. Aaron writes so well and with so much honesty you can’t do anything but falling in love with him one more time.