Google Instant Some Random Thoughts

With the launch of Google Instant I’ve got a few non-connected thoughts on what this means to SEO/SEM as well as to Google, and some questions with how they will adapt to this.

First let’s look at the comments I have specifically to Google.

A Great Engineering Feat

In one way I think this is a great engineering feat from Google, not because they are showing user’s results but because of how quickly they return the result sets. I suspect they are doing several things to take a load off the Google index. At the speed at which the results are rendered I almost have to wonder if some of these pages are pre-cached. I’m sure they could be pulling from a database new data every time a letter is pushed on a key board, but I suspect they have at the very least a set of pages (a very large set) with pre-determined content. For example there are only 26 letters in the alphabet. The second one of those letters is pressed you’ve got one of 26 pages this page can be the same for everyone, as well as the suggestions, and you could do this for several characters deep of all the permutations, even if this is several million web pages it would likely be quicker to cache the page then pull from a database, especially as it seems the suggestion do have limitations typing “lab” and “labr” don’t change suggestions until you hit “labra” to get “Labrador Retriever” a couple letters can make all the difference, as well the number of suggestions has been cut down to 5 where it was 10 before (I believe). So off the bat you cache those 26 pages, as search patterns emerge I would suspect they are taking the top search query’s and caching those pages to make returning results quicker and not needing to go back and query a database every time.

What I do think is fascinating though is when I log into my Google account, the search results change to my personalized results, in this particular case it is simply amazing at how quickly those results are coming back. I doubt there is any caching here, as these results are clearly specific to my past search history when compared to my non-logged in results, and results are still returned in less than a second. This is the engineering feat I marvel at, or maybe they just have a lot more computing power than we can imaging running this. Either way the speed results are returned is uncanny.

How Will Suggestions Scale?

Now my question is how will this scale. If the idea is to suggest what most everyone else has also searched for, how do you know when your suggestion isn’t influence a new trend? It’s kind of the lemming mentality here. If you search for “car” and everyone else is searching for “cars” does this suddenly mean you’re also going to search for “cars”? Is the search index really that much better on “cars” than it is “car”? How many users need to change a search trend before it’s reflected in the suggestions? I fear in a way the lemming mentality may be the biggest down fall Google Instant. However this is a good segue into SEO/SEM

Predictive Search

Because Google now offers predictive search, and changes results it also helps you decide which keywords to focus more closely on now. If you weren’t sure about “book” or “books” well now you’ll know it is “books”. I would imagine that most of the people will follow and go with the suggestion “books” or not even realize it is what is suggested but that they now have relevant search results. At the very least the suggestion is to go with the term that is stemmed out first from the search query than the later, as the first is likely to be the most popular now by a dramatic amount. In the pre Google Instant search days you would have done some keyword research to find that “books” was more popular than “book” but they may have been very close in search volumes I would suggest now that the disparity between the two will grow significantly.

Non-Relevant Bidding

I’ll also be interested to see if/when research comes out to show what sort of impact these flashes of other non-related search results have on a users experience. Does buying terms on non-related words ahead of your keywords impact the likelihood of a future click. Much as there is data to show a person doesn’t have to read an ad in a magazine, the act of just flipping past it can influence their behavior. I wonder if the same will hold true for search. On the plus side it may not cost you much to bid on terms ahead of yours as you only pay for clicks, on the down side these terms may be expensive as they are not relevant to your destination page. If this becomes a trend, then it also sucks if you have very short targeted keywords that are the root of other more high demand terms if people start to bid more aggressively on these earlier search terms.

Lastly bidding on single letters may become a trend. The first letter is almost like the first result now, if you can cut a person off at the first letter before they see a page that your DIRECT COMPETITOR is on then you may be able to reap in even more clicks and traffic, as you’d likely be competing with more non-related companies than related companies, again though cost of these terms could become a factor to deter this practice. Only time will tell though if it’s better to fight your competitors on the same battle ground, or pay more to fight on a field where you have few enemies.

**EDIT: Seems like Google has posted some info on the engineering feat here.  Read more at: http://googleblog.blogspot.com/2010/09/google-instant-behind-scenes.html