comScore Figures Show First Google Decline For Nearly A Year, But What To Believe?

The latest search engine share figures from comScore are now out, and Google's nearly year-long continued rises have came to a halt in July 2006, according to comScore. But how much can you trust any of the figures that ratings services provide? In this post, I look at the latest comScore stats and begin a series about how to critically evaluate search share ratings.

I've been dealing with search ratings for 10 years now. I've seen all types of things in that time. I've watched as ratings services would suddenly pull their publicly reported information, only to resume it when a competitor came along. I've watched things like AltaVista get classified as a portal rather than a search engine or important subdomains of a particular search engine not get counted. I've been bemused at how a search engine will heartily endorse figures from a ratings service when they are positive then tell investors not to trust the figures when they go against them.

My best overall advice to anyone looking at these figures is:

  • Look for long-term trends. You want to view stats for several months in a row, not two isolated months compared to each other. Stats can and will plunge from one month to the next for all types of reasons, not the least due to a ratings service itself having some counting glitch. Similarly, comparing back from one month to the same time the previous year might not reflect counting changes that may have happened or been refined over that time. I want a trend line -- and a long one.
  • Look at figures from multiple services. For several months, comScore has painted a pretty bleak figure for Yahoo, showing share decreases. At least twice this year (January & July), Yahoo has had to warn analysts not to trust the comScore figures too much (and oh the irony of Yahoo now having hired the former comScore CEO this month). In contrast, NetRatings was showing Yahoo as pretty stable. If I'm going to declare Yahoo in trouble (and I didn't), I'm more likely to do that if more than one ratings service is reflecting a plunge of some time. If it's only one of them, then I'm more in "watch and see" mode.

Now let's jump into today's figures out of comScore, which are for share of all searches within the United States:

July 2006


Searches (Millions)

Searches Per Day (Millions)

























NOTE: My figures for searches will vary slightly from comScore's as I'm working off the overall rounded 6.3 billion searches figure they provide multiplied by share percentages. I have to do that to calculate the share of "others" searches, which comScore does not provide. The comScore press release has precise figures.

Google's still shown as well above the rest. But the press release headline from comScore highlights a Yahoo victory:

Yahoo! Sites Register a Moderate Share Gain for the Second Consecutive Month

Indeed, some good news for Yahoo out of comScore for once. They have a gain that seems to come off of Google's loss.

Let's look at a big trend chart, then we'll do a drill down. Pulling the chart from my now updated comScore Media Metrix Search Engine Ratings page, we get this:

Who Gained & Lost In June & July

You can see how comScore's shown Yahoo having slight drops, then a bigger drop in January, then rises in June and July. In particular, if you go back to May 2006, Yahoo had a 28 percent share. In June, it rose to 28.5 percent. In July, a further rise to 28.8 percent.

What's fueling that gain? Maybe Yahoo Answers being popular and getting heavy promotion? Maybe something else? The answer is, we don't really know. But we can try to see who is losing.

Is Google the loser? Not in June. When Yahoo saw a gain in June, so did Google. Google went up from 44.2 percent in May to 44.7 percent. in June In contrast, MSN dropped from 13.1 percent to 12.8 percent and Ask dropped from 5.3 percent to 5.1 percent. Those drops fueled some of the gains for both Yahoo and Google.

AOL is the mystery player. comScore gave no figure for May 2006. Actually, it did -- 6.7 percent. But the following month, it clawed back that figure, saying:

Due to a definitional change occurring with June 2006 data, trended data for the Time-Warner Network are not available

Since we don't have a May 2006 figure, we can't tell if the June 2006 figure for AOL -- 5.6 percent -- was a gain from or loss to Google and Yahoo.

I've been through this type of thing before. Usually, it means that methodology has changed so much that the ratings service doesn't feel comfortable comparing figures under a new system to the old one. This note is unusual in that it's particular to AOL, rather than the methodology over all. It's also disturbing, because it suggests that all those months of AOL figures previously reported perhaps can't be trusted.

I'm asking comScore about this and will update. But let's move ahead to July. Again, the goal is to figure out who lost to Yahoo's gain. In July, it really was Google, mostly. Google dropped a full percentage point, 44.7 percent in June to 43.7 percent in July. Yahoo got some of that, a 0.3 percent rise to 28.8 percent. AOL got a 0.3 percent rise as well, coming up to 5.9 percent. The Ask network of sites also came up 0.3 percent, to 5.4 percent. MSN didn't budge.

Crisis for Google? Way, way too early to be saying stuff like that. As I said, I want to see several months of trending data from a particular player before I start issuing panic calls. For all I know, next month comScore will quietly reissue these figures that shows Google doing better.

Revising Data After The Fact

For instance, let's go back to May 2006. Did you know that a month after comScore released figures, they revised those? There was no big press release about it. It was an asterisk mention as part of the June 2006 figures.

What was different? Let's compare:

































Aside from AOL, there was no really big change among the players. But then again, this revision suddenly changed MSN from a fifth month in a row of either decreases or no gain to the different story of the first gain in five months.

MySpace The Search Monster?

Also notice how MySpace disappears off the chart, in the revised figures. What happened? Why did comScore stop reporting figures there? I'm checking on that as well, but my guess is that close attention to the MySpace figures may have backfired on comScore.

Let's go back to April 2006. That's the first time that out of the blue, we get search figures for MySpace. comScore told us: has been added to the search engine rankings for April 2006, coming in at 6th place with 43 million search queries performed (0.6 percent share of the U.S. search market). Will this smaller player eventually be able to grab a substantial share of the search market due to the site's remarkable popularity?

That's press release bait, OK? That's a nugget you're putting out in hopes that the press will eat it up. And they do, eventually. For example, in MySpace, The 27.4 Billion Pound Gorilla over at TechCrunch, the article notes at the end:

MySpace also has the sixth largest market share among search engines, even though they aren't, actually, a search engine.

I'll get to another more important press reference in a moment, but let me diverge on the "MySpace As Search Engine" idea. What's comScore counting as searches? From the May 2006 release:

qSearch includes Web searches originating from the search engines reported, other Web-based searches such as News and Image searches and channel searches conducted on portal sites (e.g., Finance and Movies). qSearch does not include Yellow Pages or Maps searches.

OK, people on MySpace searching for other MySpacers might be doing what's considered a "channel search." But that activity is extremely unlikely to translate into a search behavior that sends people to an external site or to an ad. It's simply a site search within MySpace. If that's the traffic being counted, then searches at a place like Microsoft for software support might warrant it a place on the list (and FYI, these exact types of searches have indeed gotten Microsoft mistakenly on these types of "popular search engines" lists in the past).

MySpace does allow people to search the web. Perhaps the searches counted really are only web searches (and I'm checking on this). If so, I'm still surprise to see the sudden emergence.

Why? Because of that 2.4 percent of traffic going to "others" in May 2006. Who are these others? Among them will be players like Infospace or Earthlink. For June 2006, I have NetRatings figures giving Dogpile (an Infospace property) a 0.5 percent share. That's NetRatings, not comScore, and for a different month as well. Still, it's really unusual for a new player like MySpace to just leap out of the blue like this over a more established lower tier player. When I've seen this in the past, it's simply been because suddenly, a metrics company has changed a definition and decided to count some new player that they previously ignored.

Back to MySpace. Last month, BusinessWeek cut loose with a big article about who it was seeking a search suitor, since it was such a big search powerhouse:

MySpace already drives a huge amount of traffic to search engines. It generates 100 million searches a month. In fact, 5% of all searches on the Web and 8% of all searches on Google are originated by people who come directly from MySpace.

Wow, pretty stunning numbers. And figures that frankly, I didn't believe. As I wrote in my review:

The story also gives new, amazing stats that MySpace generates 5 percent of all searches on the web. Hmm. Just a month ago, this was said to be 0.6% of all searches in the US, according to comScore. And 8 percent of all searches on Google come via MySpace? I'm checking with the BusinessWeek author, because those stats just don't make sense.

I did email the reporter, but I never heard back. Perhaps the email went astray (I mean that in all seriousness -- email does get lost or stuck in spam filters). But five percent of all searches on the web -- eight percent of all Google searches -- those are incredible numbers, if true. But they seem so untrue when you look at the 100 million figures per month stat that BusinessWeek also gives us.

The Ask network -- with a 5 percent share of just US searches -- generates 340 million searchers per month according to comScore (see the chart above). If MySpace really had 5 percent of ALL SEARCHES WEB-WIDE, then it should have far, far more than 300 million searches per month, much less 100 million.

I don't know that the BusinessWeek story came off comScore's press release bait. Heck, the stats aren't sourced at all. But when the figures start to fly and don't make sense, that's a good time for a ratings service to stop releasing them. The BusinessWeek figures weren't adding up, and perhaps that's one reason why comScore decided that MySpace no longer deserved the limelight. After listing it as a top search engine for April and May, from June onward, it fell back into the unitemized "Other" category.

I really like BusinessWeek, by the way, but this is the second time recently that I've also seen them do stumbles with figures. Many people couldn't believe the Valley Boys write-up where BusinessWeek gave Digg a $60 million valuation. Hey, I couldn't get past this section:

Digg's stature changed dramatically that day. It is now the 24th-most popular Web site in the U.S., nipping at the New York Times' (No. 19) and easily beating Fox News (No. 62), according to industry tracker

Alexa? Alexa! BusinessWeek is going to use Alexa to tell us how popular a site is? Alexa, the toolbar that so many say is so easily manipulated? Honestly, I couldn't' believe it. In this story, I'm poking at comScore -- and later this week, I'll poke at Hitwise and NetRatings. But I won't even waste my time on looking at Alexa's data, since I long ago wrote it off as being useful. But if you don't believe me, check out Matt Cutts and his A Word About Metrics, Part II post. Why, if Digg can be so popular, so can Matt. He's almost as popular as Ask. Shortly, Barry Diller will be looking to acquire him.

MySpace The Google Enabler?

comScore's not the only one to play the MySpace press release bait game. Hitwise has been having fun with it, as well. Note what Om Malik wrote about why Google felt it had to partner with MySpace:

According to data collected by Hitwise, an Internet traffic tracking service, nearly 10.8% of Google's traffic was coming from for the week ending July 29, 2006. Had Fox gone with Yahoo or Microsoft, it could have been a serious blow to Google.

Now let's step back a minute. MySpace has shot up in traffic over time, but until recently, I never heard any connection suggesting that Google was somehow gaining because of that popularity. Google was popular because it was Google and generated its own weather. Now, suddenly, Google might potentially lose 10 percent of its traffic to its site?

That's especially amazing to digest, considering that Google had no search partnership with MySpace at the time. The web search box at MySpace was (and still is for the moment, despite the Google-MySpace deal) powered by Yahoo. If anyone would lose this "massive" amount of traffic, it should be Yahoo. And since Yahoo clearly didn't fight tooth and nail for MySpace, I don't think that search box was generating much.

So how's Google getting all this supposed traffic from MySpace. I don't know -- and Hitwise doesn't seem to know, either. Let's go to the Hitwise post on MySpace traffic: MySpace and Google: What do the Numbers Say?:

In January of this year, Google received a little over 4% of their traffic from MySpace in our U.S. sample. As of the week ending July 29th, that percentage has grown to almost 11%, making MySpace the #1 upstream site for Google.

Since over the last year there has been no formal relationship between the two companies, this high volume of traffic flowing from MySpace to Google would most likely be the result of users using their Google toolbar or manually navigating to Google while engaged in their MySpace session. It would seem that this represents a great opportunity for Google to monetize a portion of that audience through AdSense listings and increase search traffic through greater exposure on MySpace.

To be clear, all Hitwise knows is that for the ISP data is monitors, it sees that a bunch of people who leave MySpace go to Google next. It doesn't know why but can only guess that people for some reason decide to go to Google after MySpace.

If those people change their minds, would that hurt Google? Maybe -- but then again, it's not 10 percent of ALL Google's traffic, as Om wrote. It's 10 percent of the US audience. Google's not given out country-by-country breakdowns that I've seen for some time, but in the past, they'd stressed that half their traffic came from outside the US. So that would reduce the blow. What reduces the blow even further is the fact that if people are doing this out of habit, they aren't likely to break that habit regardless of who ultimately got the MySpace deal.

Poking At The Other Metrics Services

Enough of MySpace -- let's get back to the Google death plunge. One of the other things I said to look at was how the other ratings services are viewing them. For now, I'm going to give you the highlights of what Hitwise had to say about July 2006:

As you can see, the top engines combined account for over 94% of all search volume in the US. Since we last released these numbers in June 2006, Google ( has increased its share from 59.3% to 60.2%, Yahoo! Search ( has increased from 22.0% to 22.5% and MSN Search ( has decreased slightly from 12.1% to 11.8%.

As you can see, Hitwise is finding that all is good with Google plus supporting the Yahoo gain that comScore also sees. So is Google in trouble? I'd say it's way, way too early to predict anything like that.

Tomorrow, I'll take a closer look at Hitwise, poke some fingers into the service and do some trending data. On Wednesday, I'll spend time with NetRatings. Then hopefully on Thursday, I'm going to drag out my special patented three-on-one ratings chart that I haven't used since the days of NetRatings, comScore and RelevantKnowledge to help you understand how to tap into multiple services to see the trends you might want to believe.

Postscript: I sent comScore some follow-up questions, and here's what I received back.

Q. General Comments?

We agree with your assessment that a single-month decline does not constitute a trend. In fact, comScore also observed a similar seasonal decline for Google during the same period last year. Fewer work days, more vacations, and reliance on academia could all contribute to Google's core user group showing lower online activity and conducting fewer searches during this time period. Bottom line – Google could be more impacted by seasonality than other engines.

Q. What happened to AOL figures in May? Am I correct that you changed your methodology and so earlier figures are no longer comparable to recent ones. If so, what exactly changed?

In June, there was a dictionary change for AOL that contributed to some of the M/M decline in Time Warner searches in the US (the dictionary defines how the Time Warner property breaks out its sites into titles, channels and sub-channels). This change does not materially alter the year-over-year market share decline that Time Warner has experienced (down nearly 3 full share points from May 05 to May 06, the last full month before the dictionary change).

Q. What happened to MySpace? You started reporting it in April 2006 out of the blue, then by June, you stopped. Was MySpace traffic being counted before April 2006, or is that the first month you looked at searches there and counted them into the searches pie? Why did they stop being broken out in June? And were they really generating more searches than someone like the Infospace network? Finally, what searches are being counted -- even those internal to MySpace to find MySpacers?

MySpace is still included in qSearch market data, we've simply chosen not to report it on our monthly market share press releases. We typically issue market share data to the press for the top 5 search sites, and simply added MySpace for a couple of months to highlight to the market that we now have the site under measurement for Web search activity. In fact, comScore data show MySpace search activity trending very nicely over the last few months.

To clarify the specifics of our MySpace reporting, we provide data for Web search only; searches done on MySpace content (e.g., user profiles, video) are not include in the MySpace totals reported in qSearch.

Q. I notice you aren't counting yellow pages or maps searches. But Google considers any local search to be part of Google Maps. Does that mean all this traffic is getting omitted now?

We do not include map-only searches in qSearch or any of our other search tracking services. There is the initial question of whether a map request actually constitutes a search, but ultimately, low marketplace demand is the primary reason for our not reporting map searches. We have, instead, chosen to concentrate our efforts and resources in the areas of greatest importance to our major search clients (i.e., all of the major search engines).

However, maps that are accompanied by business listings (as you would find in Google Maps, if you searched for a business name category) are included in comScore's Internet Yellow Pages (IYP) Search Share report. This report includes the leading IYP sites like, plus the YP/business list search activity taking place on the portals. In other words, if the result is just a map, then comScore does not report that activity in any of its search reports; if the map includes or is part of Yellow Page listings, that activity is reported in our IYP Search Share report.