Google search share - skewed by rank checkers?

[Savanadry]

You know these figures we hear of US and UK search share - how are these verified? I'm asking because I had a site recover last December after dropping out for 6 months, although it regained the ranks it never regained the volume.

We did a comparison and traffic is down year over year for the same position in the serp's. Bing / Yahoo has consistently grown over the last 2 years.

This makes me wonder, what with all the SEO's / site owners frantically checking keywords rankings daily - are the results skewed? Are all those rank trackers included? We, as webmasters, don't tend to do this with Bing, they are mostly stable so we don't need to check as often. They also don't have massive updates which ignite a frantic round of keyword checking throughout the industry.

Google would have no reason to let us know 'well actually 30% of our searches are the same people bulk searching the same phrases every day, we're actually down 20% on real searches'. They would be even less likely to let us know that of the remaining searches 50% were 'lazy searchers' - people who know where they are going but type it into the google search box because it's on their home page.

This is not meant to be a conspiracy theory, it's just what is reported doesn't seem to tally up with my personal experience. Has anyone else thought / noticed this?

[FranticFish]

I never found the numbers very good even years ago. At the lower end they can miss demand that is small but significant for a niche SME, and at the higher end I've seen up to tenfold exaggeration.

AFAIK they're the best numbers available, but that's not saying much.

I always try to run a tightly controlled AdWords campaign for a month to get real live data now. You don't always have to spend too much money if your aim is to gather impression data.

[mrengine]

It's more likely that the paid ads, images and other display blocks Google is now using above the fold has pushed your organic listing further down the page.

[mcskoufis]

I agree with all of your concerns, stats is my passion and I always want to ensure my sources are accurate when doing such research.

I don't trust 100% what comScore or Experian say about site popularity. Even though I am not aware of comScore's methodology, Experian when it was HitWise it was getting ISP data, thus counting visits to Google from a very reliable source.

Nowadays we are in a multidevice world. People use desktops or laptops at work, smartphones throughout the day, tablets at home in the evening etc.

When they say Google was the top mobile site and Facebook the top mobile app, they fail to determine how many FB app users actually go into their account via the other devices, like on a desktop in the office.

Same with the hype with mobile & tablets. I am an old-fashioned geek, I need my MS natural keyboard even on my laptop (and a USB mouse attached). Typing text in mobile or tablet or even my laptop's keyboard is totally uncomfortable for me and I wouldn't be surprised if that was the case for many others. I don't trust that mobile is the future at all IMHO. Mobile is not practical for web usage.

So I tend to make estimates based on such stats and a few months down the road compare those with the actual traffic in analytics. You will notice many discrepancies between what comscore was saying and what happened in reality.

What about households which both parents and their children all go online via the same device? Or couples using the same computer for Facebooking? Can they distinguish those? NO!

Same in social analytics which get data from Facebook's API. In Greece we have many thousands of students who go to UK for university but their profile says Greece. Does Facebook count those as Greek or UK users? Why should I trust your fancy tool which is based on such wrong data?

FranticFish's method is what I always advise to my clients.

[brotherhood of LAN]

Glad you mentioned some sources mcskoufis, as there are a number of sources that provide such info, including the search engines themselves.

Was there anywhere in particular you were referring to Savanadry?

There's an easy way to test each source. Search for a unique string 50 or so times a month and see if they pop up in any of their stats. Try using a command line tool like curl or wget, and try with a browser.

Me personally, I'd only log searches via hits to an image on the search results... avoiding most bot activity. There are plenty of browser automation tools out there though... and there isn't any solid line between bot and 'real' (human/genuine) searches, but I'd tend to trust data from ISPs.

It's worth finding out how search engines arrive at their very own figures. Maybe there's an answer buried in their FAQ's.

[mcskoufis]

Was there anywhere in particular you were referring to Savanadry?

brotherhood of LAN no not really... I am saying that I don't trust any of them and that these metrics should only be used as an indication and not as a true fact. I am a Computer Scientist & Maths person so my logic is different than theirs.

There's an easy way to test each source. Search for a unique string 50 or so times a month and see if they pop up in any of their stats.

If you are refering to Google's ex-keyword tool or its replacement, still Google has said they normalize this data. Normalization or canonicalization as it is called (if I remember correctly) in relational databases (RDBMS) theory is not the real number of searches.

So if in one month you do 50 searches for a keyword noone searches, it is quite possible to see 0 searches the next month.

That is because it is an average based on the past 12 months (then normalized before it gets presented to the end user). If you did 50 searches a month for 6 months then you would start to see data...

This is such a big issue and I am so glad we have touched it! Kudos to Savanadry!

The companies I mention charge vast amounts of money. Very few in my SEO times in UK until 2008 had access to such data, was like over $100,000 to get in...
Not sure what is the situation now, as in Greece we have NO data at all and Google has over 85% market share. Just polls from 1000-2000 people... Mental!

Would love to hear from anyone any better alternative though!

[martinibuster]

AdWords impression data.