This past Friday, as reported by GigaOm, VentureBeat, and many other sites, Google.com and several Google services went down from 3:51 pm–3:57 pm PST, just over 5 minutes. The majority of media coverage on Google being down cited a 40% decrease in web traffic. Given our web-wide visibility and measurement of well more than 10 billion web events every day, we are a good proxy for the Internet and thought we should contribute to the conversation.
According to our logs and relatively large footprint, during the time that Google was down, our servers saw 22% less web traffic than the preceding or following 10 minutes.
When we look at referrals from Google to properties directly measured by Quantcast, we saw that the number was not 0. The number dropped by 69%, meaning we still saw referrals from some Google properties. If we extrapolate what would happen if Google’s referral’s dropped to zero, we predict a momentary 32% drop in Internet traffic worldwide.
How should we interpret this data? It is important to remember that the measured drop was the short-term result of people unable to search for the sites they wanted on the search engine they are accustomed to. While many in the media, such as CNET, suggest this indicates a dangerous reliance on Google, we think it mostly reflects a habitual preference for Google. Unlike the definitive outages caused by infrastructural damage, such as the one caused by Hurricane Sandy that we described here, we expect the impact of a Google outage to ebb rapidly as users alter their browsing behavior. Bing or another search engine will begin to fill the gap and the Internet will resume functioning normally.
Sure enough, when we examined the Bing.com referrals to sites we measure, we saw a 40% spike in Bing usage that begins shortly after Google went down* and lasts for a few minutes after. While Bing searches did not come close to making up for the Google volume, the trend is clear—the Internet adapts.
Figure 2. Traffic arriving from Bing.com to sites measured directly by Quantcast on Friday, August 16th
*We suspect that the initial dip in Bing activity was caused by some proportion of Bing users that arrive at Bing via Google properties.
Posted by Jon Katz, Product Manager, and Michael Kamprath, VP of Engineering