How Google Instant affects advertisers
Google announced today the rollout of Google Instant. Now search results are shown as you type instead of waiting for you to submit a full query. This forces us to rethink the concept of an impression and suggests some new directions in ad targeting.
Google Instant changes both natural results and ads as what you type provides Google with a different idea of what you’re searching for. To quote Google:
For example, if someone types “flow” into Google.com, our algorithms predict that the user is searching for “flowers” (the predicted search) and therefore display both search results and ads for “flowers”. However, if that user then adds the letter “c” to the query, our algorithms may predict that the user is searching for “flowchart” and show the corresponding natural and paid results for flowchart.
While this doesn’t affect how results or ranked (natural or ads), it changes the meaning of an ad impression since ads may be shown for a few seconds or keystrokes while Google adjusts the displayed ads to match the predicted query. As an advertiser, does this mean they will count these “transient views” as impressions? While these fit a literal definition, they could significantly penalize advertisers whose ads show on shorter queries that are simply “en route” to more elaborate queries. Thankfully they’ve already thought this through to have a good starting solution:
With Google Instant, an impression is counted if a user takes an action to choose a query (for example, presses the Enter key or clicks the Search button), clicks a link on the results page, or stops typing for three or more seconds.
This definition of impression eliminates much of the transient effect, but not all of it, and we also have a problem interpreting the keywords reported in AdWords when a predictive search is made.
Changes to Keyword Usage
Predictive text has already been available for a while within the search box, though now it has more impact because the results change accordingly. This is likely to change the exact keywords people are likely to search. As they see suggestions, a certain portion of them will align their search with predictions. The effect is “mainstreaming” of keywords: reducing “long tail” usage to the variations suggested by Google, presumably the most common ones. If this effect is significant, it will make the most popular keywords more so at the expense of other variations, increasing competition among advertisers. This “mainstreaming” could thus effectively penalize long tail advertising strategies and drive up overall advertiser costs (and revenue for Google).
Interrupted Search Changes Intent
Suppose someone types “flo” on their way to a search for flowcharts, but because of the predictive behavior of Google Instant they see an ad for flowers that intrigues them enough to click on it. This interrupts the path / behavior of our searcher, taking more people to flowers than would otherwise have searched for flowers. This is another form of the “mainstreaming” we described above, but now we’re talking about entire topics, not just keyword variations.
Keyword Reporting of Predictive Results
Let us again consider someone who types “flo” and then clicks on an ad for flowers (note that as of this writing, “flo” was predicted as “Florida”, but assume it predicts “flowers”). What does the advertiser see as the keyword that generated the click? The “right” answer could be either “flo” or “flowers”. They typed the former, but they responded to ads shown for the latter. AdWords is unlikely to report both, so one way or another we have only learned half of the story: either we know what they actually typed (very useful, but may not make it sufficiently clear what ad they saw and hard to use for keyword targeting) or we think they typed the whole keyword when that might not have been what they were going to look for.
It’s impossible to know at this time how important these factors will be. Most people may ignore the predictive features and changing search results entirely. Or a lot of people could become very good at using predictive text instead of typing entire queries, just like text messaging is replete with shorthand notation for common phrases (LOL, TTYL). Will serious shoppers favor full queries while casual searchers will be more likely to use predictive results? As advertisers, will we be able to tell which is which? We look forward to answers.
Although Google Instant doesn’t change the way ads are served, ads and search results will now be shown based on the “predicted search.” For example, if someone types “flow” into Google.com, our algorithms predict that the user is searching for “flowers” (the predicted search) and therefore display both search results and ads for “flowers”. However, if that user then adds the letter “c” to the query, our algorithms may predict that the user is searching for “flowchart” and show the corresponding natural and paid results for flowchart.Although Google Instant doesn’t change the way ads are served, ads and search results will now be shown based on the “predicted search.” For example, if someone types “flow” into Google.com, our algorithms predict that the user is searching for “flowers” (the predicted search) and therefore display both search results and ads for “flowers”. However, if that user then adds the letter “c” to the query, our algorithms may predict that the user is searching for “flowchart” and show the corresponding natural and paid results for flowchart.
September 26th, 2010 at 7:42 am
It has been quite few weeks that Google announced its Instant feature. Here is a poll on how advertisers feel about it.
http://www.seroundtable.com/archives/022942.html
September 30th, 2010 at 1:57 pm
I honestly don’t think that it will have that big of a affect on search results. People already have a search term in mind when they are typing in what they are looking for. I myself just type in the search term really quick without even looking at the instant results. But then again I am a pretty quick typist.
September 30th, 2010 at 2:21 pm
SEOmoz recently published some data showing it has little impact on NATURAL search traffic: http://www.seomoz.org/blog/google-instant-fewer-changes-to-seo-than-the-average-algo-update