Jumpshot's data here is only for the US. #4: What percent of Google searches result in a click? 66% of distinct search queries resulted in one or more clicks on Google's results. That number is 0.9% of Google search clicks, just under 1 in 100. #11: What percent of clicks on Google search results go to YouTube? Jumpshot, however, thanks to clickstream tracking, can see that 0.16% of search clicks go to Gmail or Google Mail following a query, only a little under the number of clicks to tweets. #13: What percent of clicks on Google search results go to Google Shopping results? Honestly, I'd have estimated this in the 20–30 percent range, so it surprised me to see that, from Jumpshot's data, all Google properties earned only 11.8% of clicks from distinct searches (only 8.4% across all searches). You search. Jumpshot: No.
One of the marketing world’s greatest frustrations has long been the lack of data from Google and other search engines about the behavior of users on their platforms. Occasionally, Google will divulge a nugget of bland, hard-to-interpret information about how they process more than X billion queries, or how many videos were uploaded to YouTube, or how many people have found travel information on Google in the last year. But these numbers aren’t specific enough, well-sourced enough, nor do they provide enough detail to be truly useful for all the applications we have.
Marketers need to know things like: How many searches happen each month across various platforms? Is Google losing market share to Amazon? Are people really starting more searches on YouTube than Bing? Is Google Images more or less popular than Google News? What percent of queries are phrased as questions? How many words are in the average query? Is it more or less on mobile?
These kinds of specifics help us know where to put our efforts, how to sell our managers, teams, and clients on SEO investments, and, when we have this data over time, we can truly understand how this industry that shapes our livelihoods is changing. Until now, this data has been somewhere between hard and impossible to estimate. But, thanks to clickstream data providers like Jumpshot (which helps power Moz’s Keyword Explorer and many of our keyword-based metrics in Pro), we can get around Google’s secrecy and see the data for ourselves!
Over the last 6 months, Russ Jones and I have been working with Jumpshot’s Randy Antin, who’s been absolutely amazing — answering our questions late at night, digging in with his team to get the numbers, and patiently waiting while Russ runs fancy T-Distributions on large datasets to make sure our estimates are as accurate as possible. If you need clickstream data of any kind, I can’t recommend them enough.
If you’re wondering, “Wait… I think I know what clickstream data is, but you should probably tell me, Rand, just so I know that you know,” OK. 🙂 Clickstream monitoring means Jumpshot (and other companies like them — SimilarWeb, Clickstre.am, etc.) have software on the device that records all the pages visited in a browser session. They anonymize and aggregate this data (don’t worry, your searches and visits are not tied to you or to your device), then make parts of it available for research or use in products or through APIs. They’re not crawling Google or any other sites, but rather seeing the precise behavior of devices as people use them to surf or search the Internet.
Clickstream data is awesomely powerful, but when it comes to estimating searcher behavior, we need scale. Thankfully, Jumpshot can deliver here, too. Their US panel of Internet users is in the millions (they don’t disclose exact size, but it’s between 2–10) so we can trust these numbers to reliably paint a representative picture. That said, there may still be biases in the data — it could be that certain demographics of Internet users are more or less likely to be in Jumpshot’s panel, their mobile data is limited to Android (no iOS), and we know that some alternative kinds of searches aren’t captured by their methodology**. Still, there’s amazing stuff here, and it’s vastly more than we’ve been able to get any other way, so let’s dive in.
23 Search Behavior Stats
Methodology: All of the data was collected from Jumpshot’s multi-million user panel in October 2016. T-distribution scaling was applied to validate the estimates of overall searches across platforms. All other data is expressed as percentages. Jumpshot’s panel includes mobile and desktop devices in similar proportions, though no devices are iOS, so users on Macs, iPhones, and iPads are not included.
#1: How many searches are *really* performed on Google.com each month?
On the devices and types of queries Jumpshot can analyze, there were an average of 3.4 searches/day/searcher. Using the T-Distribution scaling analysis on various sample set sizes of Jumpshot’s data, Russ estimated that the most likely reality is that between 40–60 billion searches happen on Google.com in the US each month.
Here’s more detail from Russ himself:
“…All of the graphs are non-linear in shape, which indicates that as the samples get bigger we are approaching correct numbers but not in a simple % relationship… I have given 3 variations based on the estimated number of searches you think happen in the US annually. I have seen wildly different estimates from 20 billion to 100 billion, so I gave a couple of options. My gut is to go with the 40 billion numbers, especially since once we reach the 100MM line for 40 and 60B, there is little to no increase for 1 billion keywords, which would indicate we have reached a point where each new keyword is searched just 1 time.”
How does that compare to numbers Google’s given? Well, in May of 2016, Google told Search Engine Land they “processed at least 2 trillion searches per year.” Using our Jumpshot-based estimates, and assuming October of 2016 was a reasonably average month for search demand, we’d get to 480–720 billion annual searches. That’s less than half of what Google claims, but Google’s number is WORLDWIDE! Jumpshot’s data here is only for the US. This suggests that, as Danny Sullivan pointed out in the SELand article, Google could well be handling much, much more than 2 trillion annual searches.
Note that we believe our 40–60 billion/month number is actually too low. Why? Voice searches, searches in the Google app and Google Home, higher search use on iOS (all four of which Jumpshot can’t measure), October could be a lower-than-average month, some kinds of search partnerships, and automated searches that aren’t coming from human beings on their devices could all mean our numbers are undercounting Google’s actual US search traffic. In the future, we’ll be able to measure interesting things like growth or shrinkage of search demand as we compare October 2016 vs other months.
#2: How long is the average Google search session?
Form the time of the initial query to the loading of the search results page and the selection of any results, plus any back button clicks to those SERPs and selection of new results, the all-in average was just under 1 minute. If that seems long, remember that some search sessions may be upwards of an hour (like when I research all the best ryokans in Japan before planning a trip — I probably clicked 7 pages deep into the SERPs and opened 30 or more individual pages). Those long sessions are dragging up that average.
#3: What percent of users perform one or more searches on a given day?
This one blew my mind! Of the millions of active, US web users Jumpshot monitored in October 2016, only 15% performed at least one or more searches in a day. 45% performed at least one query in a week, and 68% performed one or more queries that month. To me, that says there’s still a massive amount of search growth opportunity for Google. If they can make people more addicted to and more reliant on search, as well as shape the flow of information and the needs of people toward search engines, they are likely to have a lot more room to expand searches/searcher.
#4: What percent of Google searches result in a click?
Google is answering a lot of queries themselves. From searches like “Seattle Weather,” to more complicated ones like “books by Kurt Vonnegut” or “how to remove raspberry stains?“, Google is trying to save you that click — and it looks like they’re succeeding.
66% of distinct search queries resulted in one or more clicks on Google’s results. That means 34% of searches get no clicks at all. If we look at all search queries (not just distinct ones), those numbers shift to a straight 60%/40% split. I wouldn’t be surprised to find that over time, we get closer and closer to Google solving half of search queries without a click. BTW — this is the all-in average, but I’ve broken down clicks vs. no-clicks on mobile vs. desktop in #19 below.
#5: What percent of clicks on Google search results go to AdWords/paid listings?
It’s less than I thought, but perhaps not surprising given how aggressive Google’s had to be with ad subtlety over the last few years. Of distinct search queries in Google, only 3.4% resulted in a click on an AdWords (paid) ad. If we expand that to all search queries, the number drops to 2.6%. Google’s making a massive amount of money on a small fraction of the searches that come into their engine. No wonder they need to get creative (or, perhaps more accurately, sneaky) with hiding the ad indicator in the SERPs.
#6: What percent of clicks on Google search results go to Maps/local listings?
This is not measuring searches and clicks that start directly from maps.google.com or from the Google Maps app on a mobile device. We’re talking here only about Google.com searches that result in a click on Google Maps. That number is 0.9% of Google search clicks, just under 1 in 100. We know from MozCast that local packs show up in ~15% of queries (though that may be biased by MozCast’s keyword corpus).
#7: What percent of clicks on Google search results go to links in the Knowledge Graph?
Knowledge panels are hugely popular in Google’s results — they…