On Monday Google announced that it was allowing public testing of its “next generation” infrastructure.  The architectural changes, dubbed “Google Caffeine”, aren’t just speeding up Google’s ability to process queries–searches conducted using the Caffeine sandbox are showing different results.

While the differences may be subtle, we all know that a move of one or two slots can have a significant impact on traffic.  Thus, it’s no surprise to find a number of people in the SEO world studying the differences between the test environment and traditional, “decaffeinated” Google.

At this point, publicly shared testing results are anecdotal and it would be a stretch to read that much into them.  That’s particular true when one considers the fact that Google is soliciting developer opinions about Caffeine, which means there’s still a propensity for future changes and refinements.

With that understanding in mind, here are few initial observations about “Google Caffeine”.

SPEED.  While we’re basically taking Google’s word for this (at least I don’t have the ability to time things to the microsecond accurately), it appears as though Caffeine is faster than regular Google.  Testers have experienced delivery times that are twice as fast as usual.

BLENDED CONTENT:  A few people have noticed that sandbox results have video and photo content pushed further down the page than usual.

KEYWORDS:  While you’d think that any Google refinements would place less emphasis on “exact match” keyword use, Ben Parr feels that Caffeine is placing greater weight on matching keyword use than the standard Google engine.

SOCIAL MEDIA:  Twitter, FaceBook and other social media items seem to be performing better in the sandbox.

I’ve done some preliminary testing myself and have noticed some of the same things others have.  In terms of subjectively assessing the quality of the results, however, I don’t see an appreciable difference.  I’ve noticed multiple occasions where a #4 and #5 result might flip-flop or where something may come up from #8 to #3, etc.

These changes, however, seem to happen when the likely value of the sites to the user are very similar.  I haven’t observed any foolish promotions of low-caliber sites to higher SERPs in my limited exploration of Caffeine.  The only exception is the occasional presence of Twitter-related items that don’t pack much of a punch in terms of user value creeping into the top 10.  In Caffeine’s defense, however, these seem to be displacing fairly low-quality entries in the first place.

Admittedly, my testing is anything BUT scientific and I’m nowhere close to having a valid sample size from which to reach any reliable conclusions.  Like others, I’m just relaying what I’ve found so far with a bit of tinkering.

It will be interesting to read more about what differences others are finding.  What will be even more interesting is to see whether the final roll-out produces results that are measurably different from those of the beta test.  Google is asking for input and unless they’re sitting on a well-received model, we’ll undoubtedly see additional changes prior to implementation.