Google Instant is an amazing piece of technology. However, I imagine, like most techies, the question that first springs to mind is “Oh my god, how much data is this sucking down?!?”
The answer of course is: “it depends”. It depends to a large degree on the kind of results you’re going to see, how many results there are on the page, whether they have maps in them or images, and lots of other factors. It also depends on how accurate the query suggestions are at guessing what you are going to type, since the more accurate it is the fewer times it will have to re-fetch results from the server.
For instance, let’s say I’m going to look up train timetables from Victoria Station, London. I start typing, and when I put in the first letter ‘v’ Google makes a wild guess that I’ll be looking for Verizon and grabs down results for it. So far 13.5Kb of search result data has been sucked down, an increase of just under 13Kb over the non-instant option, which just sucks down the suggested search queries, not the results themselves.
Victoria's Secret not Victoria Station
When I type the letter ‘i’, Google realises I’m not looking for Verizon and decides I must be looking for Victoria’s Secret. That adds another 29Kb to be sucked down, which includes a couple of images. (Which are pretty tame by the way, I have safe search on at work).
Now, 29Kb is pretty small. Google have compressed the data, and since search result data is very compressible it averages about a 70% bandwidth saving, good for what is essentially pure text with some images thrown into the data.
From that point until I get all the way to ‘Victoria St’, my results stay static, since it looks increasingly likely that I’m looking for lingerie. However, there is another 10K pulled down, or about 1.4Kb per keystroke. This isn’t results, just different suggestions being cycled through the list as I type, (vicodin, victoza, victor) but Google is still showing results for what it thinks is the most likely option – Victoria’s Secret.
This behaviour is the same as it is for the existing search suggestions so we’ll discount the data for that. When I’ve got to ‘Victoria St’ Google realises its embarrassing mistake and decides that I must be searching for Victoria Stilwell the famous dog trainer. That adds another 25Kb, again with images encoded into the results.
Victoria Station
When I get to ‘Victoria Sta’ the penny drops and Google gets Victoria station results, which weigh in at just 11Kb, with no images, and from then on to the end, the results don’t change, except for the cycling dance of other possible auto complete suggestions (victoria stafford, victoria station salem etc.)
In total then Google Instant added 89Kb in downloaded data over and above what a previously standard experience would have required. A tiny test of 20 other random queries from my own search history shows this to be pretty average. Obviously maps and image data which are not in the final result set add to this, but calling it 90Kb extra per search (with 6 queries in the search) seems to be in the ballpark.
This maps pretty well to Google’s own expected figures. They reckon they’ll see 5 to 7 extra search results fetched as a result of an Instant search, and presumably they know what they’re talking about. How much it is used and how accurate it will be is anyone’s guess.
Taking Google’s current round estimate of 1 billion searches per day and 6 as the midpoint of their reckoning of accuracy, and my finger-in-the-air of 15Kb for the data for each extra set of results, we get a pretty measly 85 Terabytes extra of data leaving Google’s server farms and the average UK user, who averages around 4 searches per day, getting an extra 360Kb per day down their internet connection. This is hardly a noticeable amount of data for a corporation that deals in Petabytes for its indexing of the web. Similarly, 360Kb is hardly noticeable for a user with even the slowest of broadband connections.
But is there any point? In all my use of instant so far, it’s felt like no more than a bothersome distraction. I do use Google suggest pretty often for long tail searches, and it’s easy to see if what’s being suggested describes what you want to type.
However, looking down at the search results is a further glance away, and the information takes longer to interpret. It feels unnatural to me. If I’m typing a query string, I’m typing text, so a suggestion of what I am going to type may be helpful.
On the other hand a suggestion of search results for that query isn’t what I have in my mind. It’s another step away from the thought in my brain that millisecond.
Time will tell I suppose, but if Google Instant isn’t an instant hit I’d expect to see it become opt-in rather than opt out for Google users by default pretty quickly.
How have you found using Google Instant so far – do you like it, hate it, or haven’t really noticed it at all? Share your thoughts in the comments.
Vs. 
Most Recent Comments