This would be an interesting way to choose a startup name. E-mail gathering splash-page targeted to keywords within your industry.

[Name 1] - The service that helps you [solution to problem].

[Name 2] - The service that helps you [solution to problem].

[Name 3] - The service that helps you [solution to problem].

He's doing it wrong, if you look at the impressions he's got adwords set so that if a keyword is doing well it gets more impressions. That's the wrong way to do this set, you have to configure it to give each keyword equal coverage.

Otherwise what happens is that by pure randomness some of your keywords are going to get less clicks than others, and as soon as that happens to one of your keywords they get "frozen" (i.e. they get no more ad impressions so they get stuck at their "random" low value).

I'm no detective, but here are some of the names:

  Annabel Strange
  Annabel Spring
  Annabel Sketch
  Annabel Scrape
  Annabel Start

Did you de-obfuscate the text, run the searches, or use some other mechanism? I didn't find anything when running the searches and don't really have the time to mess with de-obfuscation

I Googled "She's the Sherlock Holmes for the 21st century", and that gave me 3 pages of sites that were displaying the ad when they were indexed.

And now this page is the number 1 hit in google for that phrase and some little vector spread across racks of google boxes is wondering why hackers suddenly care so much about Annabel and what they can sell us.

LOL who would have thought Google would have indexed ask.com pages showing their own Adwords ads!

It's the ultimate form of "eating your own dogfood".

Also Annabel Slack, Annabel Stark, Annabel Query, Annabel Strake, Aanabel Streem

Tip: Google the query like this:

+annabel +"She's the Sherlock holmes for the 21st century"

The Annabel variants will all appear in the search summary. No need to click through to the cached page.

Looks like one of them is also Annabel Scheme

Edit: and the other I just found is Annabel Stymie

That is actually a pretty clever way to do quick and cheap marketing research. However, not a good way to write a novel. To all aspiring writers: please do not base your novel on marketing research, if I wanted that I would be watching TV.

If the author had A/B tested the headline, it probably would have gotten more clicks as "Naming characters in a novel". The way it reads now, it sounds like something about using non-ASCII symbols in Adwords.

People who are familiar with that definition for "characters" probably aren't the audience he was thinking of. =)

Tim Ferriss gives a good example of this in the book Four Hour Work Week, when he discusses how he used Adwords to pick the book name.

You have to be careful when doing this not to violate Google's TOS. I don't think they like people using AdSense just for "market research" because the users who actually click on the link get taken to a useless landing page. Unless he had a pre-order form for the book as the landing page, that's not generally accepted. I know Tim Ferris kind of popularized that technique, but if all of Google's Ads lead people to worthless landing pages, their CTR across the network would go down. Just an FYI for those who can't afford to get a Google Ban because they also rely on AdSense.

> Unless he had a pre-order form for the book as the landing page, that's not generally accepted.

This sounds like an assertion with no real knowledge to back it up.

What does "generally" mean?

Either it's acceptable or it's not.

Here are the adwords TOS:

http://adwords.google.com/support/aw/bin/static.py?page=guid...

This may be what you're thinking of:

http://adwords.google.com/support/aw/bin/static.py?page=guid...

Your ads and keywords must directly relate to the content on the landing page for your ad. When users see your ad, they should be able to understand what kind of product, service, or other content they will find on your site. Products or services promoted in your ad must be reflected on your landing page; ads can be disapproved if a promoted product is not offered or available for sale as promised.

The ad doesn't promise that anything is for sale - it seems like a general branding campaign.

Yeah, that must have been what I read and extrapolated a bit.

The FTC doesn't like ads implying things are for sale that are not, either -- but at this scale enforcement is unlikely.

Why do some of the ads get run more often than others? Seems like google would run the one with the highest click through rate more often. I'm not convinced all of these are statistically significant anyway, but I don't have R at the moment.

There is a setting in Adwords for "rotate my ads evenly over time" that is necessary for accurate A/B testing. By default, Google shows the ads with the highest click rate more often, so you'll get a bias.