Google’s leapfrog from search term Suggest (now renamed Autocomplete) to the new “Google Instant” predictive results service is impressive, but it would be so much more compelling if the search giant explained a little more about how its internal software development teams bring about the changes that they do.
It took years for Google to even open up a view of its data centres to the outside world, so we don’t expect the full recipe for Instant’s ‘secret sauce’ — but a little more background in terms of engineering would be welcome.
Presented yesterday by the company’s VP of search product and user experience Marissa Meyer, Google Instant is a new infrastructure enhancement designed to make search easier by showing you results right on the homepage as you type.
Google says that it estimates that Instant will save typical searchers between two and five seconds on every query — and that if every Google searcher used Instant, this would save more than 11 hours per second.
This is about as deep as this week’s announcement gets, “On the back-end, Instant is pushing the limits of our technology and infrastructure. For typical searches, we estimate we’ll show between five and seven times as many results pages as before. With Google already serving more than one billion searches each day, we needed to find a way to efficiently serve all that content.”
When asked for more information on the programmer/developer angle of this announcement, Clara Armand-Delille, corporate communications and public affairs manager for Google UK said, “With Instant, we’re introducing a couple of new features that include dynamically displayed results right on the homepage so you can quickly review and find your search result. Predictive text saves time by guessing your search so you don’t have to finish typing. Lastly, scroll to search enables you to manually scroll through a list of Autocomplete predictions with the arrow keys and instantly see results for each.”
Ah so no special sauce – so far this is about all we know, the Google algorithm is as follows…
PR(A) = (1-d) + d(PR(t1)/C(t1) + … + PR(tn)/C(tn))
PR(A) stands for the Google page rank of our arbitrary example page A
t1 – tn are the pages that link to page A
C is the number of outbound links that a page has and in this case our C variable is examining pages t1 to tn
d is a damping factor, which is usually set to 0.85 – this is a standard function used when working with numerical algorithms
It is likely that Google has progressed the current form of this equation, but that it is still largely based on this initial form. As for secret sauce, no luck this time.