This post/question was inspired by this one: Telepathy, Inc. Puts Massachusetts.com on the Market with Sedo.
How does Google count typos that it auto corrects?
I search for 'Massachusets' (typo) and it shows me results for 'Massachusetts' (proper spelling) which keyword does that count for? What about for advertisers?
Example of Auto Correction:
What we see is Google recognizes that we've probably made a typo.
Instead of showing results for the typo, it shows the results for the correct spelling.
This is fantastic for user experience, it gives better (more expected) results.
As an advertiser, I may be trying to get a better ROI by taking advantage of typos.
My decision to do this would be under the assumption that they were probably searching for
the same thing with and without the typo. There would be little/no bias in demographics.
Everyone makes typos (of course you should always test each hypothesis!).
But, if Google auto-corrects search results and it requires an extra step to search
for the typo, then it begs the questions
- who are these typographically/spelling challenged searchers?
- Are all searches auto corrected (eg. Google Search vs Domain Parking)?
I tried searching for an answer but didn't find anything.
Investigating Search Counts
The basic idea behind this methodology was, if there were wildly different ratios across various sources that correlated with auto correction being enabled, we could conclude whether or not it affected search counts.
Google ~ 30.45:1
Bing ~ 35.15:1
Bing also has some sort of auto correct going on:
Both Massachusetts.com and Massachusets.com are parked pages. So I figured both would be pretty much just getting type in traffic as nothing was developed. Does the ratio hold?
Not enough data to say anything conclusive on Compete, sadly.
Wordtracker ~ 20.63:1 (5654:274)
I tried to find the old AOL search data to see if I could match this. I couldn't remember the site which used to allow you to search this information.
Conclusion
Google and Bing both seem pretty similar, Wordtracker is substantially lower, but I have no idea how they collect their data and from where. I think their sample is smaller and take the results less seriously. I checked some other sources (old Overture data for instance on domain matches) which didn't yield anything.
The answer is, I am still not sure. Maybe Quora can shed some light?
