Contributors & Commentary
Potential Factors Not Included in the Survey
Ian Lurie discusses the trending of searches as possible factor in predicted search suggestions. If you have insights or data to add to this, please connect with us below.
Competitive in AdWords Bidding
Tim Eschenauer states that “Google knows many searchers are influenced by what they suggest, at least at the beginning stages of their search. You’d think they’d at least try to lead them to the paid results.” Share your thoughts with us on the impact of PPC search data to organic autocomplete suggestions and we will add it to our next survey.
Google admits that “you may see search queries from relevant searches that you’ve done in the past.” Many SEO experts, including contributors of this article search for their own rank frequently on terms such as “SEO Expert” and “SEO Consultant”. Even with Personalized Search turned on, autocomplete does not return predictions for either query:
If you have examples of different circumstance or testing with regards to autocomplete and Personalized Search, please let us know and we will include it in the next survey.
Special Thanks to Our Contributors:
Cayley Vos, Owner
We have seen dramatic increase in clicks through to long tail searches suggested by Google. It also allows other websites to capitalize on brand searches. We ranked for a keyword ‘brand discount’. An example search would be “Nike Discount” and this search greatly helped the smaller retailer with a 25% traffic increase.
Tom Critchlow, Head of Search
I’ve only done some testing around this subject so don’t have definitive answers but I’m a strong believer that the search suggest options are generated from Markov-chain type processes. Obviously on top of that if there is large search volume Google wants to add value by using CTR, volume of mentions etc etc.
But Google also wants to provide search suggest results for long tail queries and I believe they do this by document analysis, Markov-chaining and machine learning because I’ve seen numerous examples of search results where for example there is a “brand x sucks” search suggest but where this phrase appears nowhere online and is also not searched for therefore I believe this hints that Google are creating this from somewhere by knowing that “brand x” refers to a brand and that “brand x sucks” is a common search format (obviously this is a very basic example but demonstrates the concept). Remember that Google drove a car through a city for 100,000 miles without a driver. They know a thing or two about machine learning.
Ian Lurie, President
I think trending is also really, really important. If a particular set of phrases around a topic are trending sharply, those phrases are more likely to end up in search suggest.
Search suggest is influenced primarily by click-through and trends. If increasing numbers of people are searching on phrases and clicking through at a steady rate, and those phrases are easily clustered around a topic, you can bet they’ll end up in Search Suggest.
Rand Fishkin, CEO
I suspect the formula for it is both more and less complex than we’d suspect – less complex in that the quantity of signals may be just a few (maybe amount of use/appearance on the web, search volume and usage/CTR) and more complex, e.g. machine learning against a sample set of queries and suggestions to analyze the usefulness/effectiveness of the algorithm producing them (and whether they’re working well for users).
Search suggest is likely one of Google’s best methods for helping those stuck on what to query and reducing the uncertainty of intent. I suspect it also positively influences number of searches per searcher, particularly the number of searches where ads appear.
Aaron Wall, SEOBook Founder
In my opinion, Google autocomplete predictions mainly come down to search volume. The more a phrase is searched, the more likely it will be to appear in the suggestions.
Tim Eschenauer, SEO/SEM Strategist
My feeling is that Google Suggests terms that have high search volumes and perhaps are competitive on the PPC space. Google knows many searchers are influenced by what they suggest at least at the beginning stages of their search. You’d think they’d at least try to lead them to the paid results.
We know how busy life gets when you’re the leader in your industry. We want to thank those who participated in the study despite their crazy schedules. Not listed above, but noteworthy of participating nonetheless, are:
Danny Sullivan, Editor and Chief
Robert Wright, Search Marketing Expert
Comments & Feedback
To participate in the next study, please call us at (562) 732-4417. To comment on the first run at analyzing Google Autocomplete, please use the Facebook comment section below. Thanks for reading!