Google Instant And The Power Of Suggestion

It’s going to be a long time (weeks or months) until anyone has a complete understanding of how Google Instant will change search and search marketing.

The change Google has made isn’t just to the user interface. They’ve changed how search works by dramatically increasing the power of Google Suggestions. And most importantly they’ve changed the search experience which will undoubtedly change the way people behave and react to the results they get from Google.

Search marketers—of both the paid and organic varieties—are going to feel the impact of all these changes.

Instant suggestions

Watching a Google demo of Instant, or when giving it a casual try yourself, you can’t help but be struck by the way the screen is constantly updating and adapting as you type. But if you pay close attention to what is really happening you’ll notice that all the action on the page is driven by the fact that Google Suggest is constantly executing on its top suggestion—a suggestion that usually changes (but not always) after each new character is typed.

Last night my son needed to gather and print some articles for a homework project. Using his query as an example—a search for “marine pollution”—let’s see how Google Instant really works and begin thinking about the implications for search marketers.

For the sake of this examination, we’ll type slowly enough to allow the screen to react to every single character. In the real world people don’t type that slow, and at least in my experience Google Instant either needs some time or is engineered to avoid the pointless flashing of results. So many searchers wouldn’t see all of the steps we’re about to review.

Type “M”—Suggest = mapquest

The first suggestion is that we’re looking for MapQuest. And there isn’t even a house ad for Google Maps! I assume that Mapquest is the most popular search in the world that begins with the letter “M.”

Or is it? One would expect that these suggestions are personalized in some way, based on the user’s search history, the geography of the user, the day or time of the search, or the contents of the last few messages that landing in your Gmail account (just kidding on that last one). Does everyone get mapquest for “M”?

What I find most interesting about the way Suggest is being used is that it’s 100% focused on the first suggestion in terms of the delivered results. When this was just a suggestion list, first is first. But when actually assuming that is what I want, it seems like an overweighted bet.

The second suggestion is “myspace” and the third is “msn.” Why not serve up a result set that is a weighted mixture of the results for these top three? Wouldn’t that increase the odds that what I wanted was “instantly” available?

Type “Ma”—Suggest = mapquest

Here’s proof that it’s not character-by-character instant. Adding the “a” changes nothing. The suggestion remains “mapquest” and neither organic nor paid results change.

This makes the focus on the first suggest seem even stranger to me. Google had already served me a full page of Mapquest results, it knew I paused for longer than 3 seconds so I saw them and didn’t click but rather kept on typing. Doesn’t that suggest that their suggestion was wrong? Aren’t the odds of “myspace” or “msn” even higher now? Shouldn’t the results start hedging their bets?

Type “Mar”—Suggest = marriott

When another character proves that I don’t want mapquest, Google Instant suggests “marriott.” They obviously know that I need a vacation. Or marriott is the top “mar” search in the world.

As others have pointed out, brands do well in the world of suggestions—both Mapquest and Marriott are brands and probably two that spend a lot on AdWords, though in this case only an ad for Marriott has appeared.

The runner-up suggestions were “marshalls,” “mario,” and “marathon grill.” More brands. This is likely simply due to the click density that these type of “head terms” get as compared to the zillions of long tail queries that begin with “mar” rather than some overt brand favoritism. But the effect is the same—brands will enjoy greater impression counts and likely some collateral incremental clicks.

Type “Mari”—Suggest = mario

As I continue typing and get closer to one of the existing suggestions, Google Instant goes right for it. Paid ads appear for the first time. There doesn’t seem to be a consistent number of characters before AdWords ads appear—with “R” it’s just one but in many other cases it’s 4 or even 5.

Type “Marin”—Suggest = marines

Type “Marine”—Suggest = marines

Type “Marines_”—Suggest = marine corps

The trend continues, as more characters that support the top suggest yield no instant changes. It’s interesting to see (and a consistent behavior) that spaces do reset suggestions. So when I fail to type the “s” they’re expecting the suggestions and results are reset, although in this case to the derivative “marine corps.”

Now someone needs to come up with the clever name for winning two suggestions in a row with different keywords from the same website!

Type “Marine P”—Suggest = marine parts

As another typed letter proves suggest wrong yet again, it doggedly keeps trying. Switching to the other kind of marine, the suggestion now is that I learn about boat parts. “Marine plywood,” “marine paint,” and “marine power” are the next suggestions but again 100% of the organic and PPC results focus on “marine parts.”

At this point I’ve typed 8 characters (including the space) and been shown five different sets of results, none of which have a thing to do with the subject I’m looking for. I’m sure they’ve all been statistically sound probabilities. But as a user it feels a little bit like talking to someone who won’t let you finish a sentence even when you have not provided enough data to reasonably make any response.

In terms of search marketing, there might have been something along this path that caught my eye and got a click, but in that case I was diverted from my intent and may or may not ever go back and pick up the original trail. On the other hand, I may be tiring of all these incorrect results—in a way I may come to think that I’ve done at least 5 Google searches and they’re still not even close to what I want. It’s getting annoying and tiring. Or I might like the responsiveness and their attempts, they may learn more about me over time and get better at it, and I might double my use of Google.

I’m guessing there will be people in all three camps, and the eventual percentages are what we’re all waiting to know.

Type “Marine Po”—Suggest = marine power

After I enter this character we’re getting somewhere. Another page of incorrect results but my intended query is now #3 on the suggestion list.

Type “Marine Pol”—Suggest = marine pollution

We’ve arrived.

Google Instant serves up results that satisfy the original intent, and the paid ads to compliment that query.

Note that when I type at a more normal pace, the screen refreshed only about 3 times before I got to my desired results. And I can’t say that the results would have impacted me at all because I was going very deliberately towards my goal. Again, this test may not reflect the experience that most searchers will have in the real world. A range of experiences and reactions is probably what we can expect.

Observations

The conversion of Google Suggest into Google Presume has gigantic implications beyond the obvious. I think it means that more people will see, and likely settle for a smaller pool of results. Both those that were diversions from and those that are “close enough” to the original desired results. If this is true then tracking and becoming relevant for the words and phrases in Google Suggest just became the #1 goal of search marketers everywhere.

What’s the URL of that new site where you can upload your current keyword list and they’ll return a harvested list of every suggestion made for every character combination seen on the way to that keyword? I don’t know yet but am sure we’ll all hear about it soon.

On the other hand, users might react to the power they now have and take more control rather than taking more suggestions. They may see the presumed results but realize that it takes very little effort to keep refining the query, seeking their original goal or perhaps even experimenting around it. They may in fact get more results, better results, and extend or at least keep alive the long tail and results diversity.

This is the great unknown. How will users react and adapt their search habits. That’s the one that will undoubtedly change over the coming weeks and to which search marketers will have to react just as surely as we have to react to the UI and algorithm changes of Google Instant.

Here is a video of my entire experiment:

[youtube]https://www.youtube.com/watch?v=ln45lCncT2c[/youtube]