I've read about this method for name generation many times before and there always seems to be a lack of data to show the method's level of quality or relevance. Would love to hear a statistician's take on this.
Fully agree, but the author is not using AdWords to generate names, but to have a number of names (4 in this case) compete with each other (running the exact same ad, just changine the name). The reported outcome in that case was 80% of the clicks went to "Autovate", although we don't know that all variations got equally exposed.
[edit] The four names in question were Motiveatr, Autovate, Autovator, My Auto Motive.
Completely agree! Let's say the author got 20c clicks. So, for $20, he got 100 clicks. And these clicks were divided among 4 names. This is obviously not statistically significant data.
This concept can even be taken further - rather than testing just a name, use AdWords to test your entire product before even building it.
For example, if you are thinking of opening an online business selling widgets, you can build a test website and advertise it using AdWords. Provide a fake checkout process, but rather than processing credit cards, just inform your users that your product is not yet available (at a point in the workflow where it's clear they intend to pay already). Then, with a quick calculation on the data, you have a rough idea of how profitable your business will be upon launch.
Obviously, it's not a perfect model, but still a very nice way to gather market data for cheap before you invest the time and money to build anything.
With $20 budget and 4 versions this seems like a random spike in data. If you are going to use this method I suggest to use Google AdWords Campaign Experiments (built-in tool). It will tell you if your results are statistically significant. There is also Excel spreadsheet to download and play with numbers to see if results are significant.
Second caution. This approach only tests what's the best name for your ad. It doesn't tell you what is the best name for your product.
Interesting method but the author seems a little surprised that doing market research actually works. His sample size compared to the population seems just a liiiiittle bit too small to yield meaningful results from though and there's no real indicator of its reliability.
What would really be interesting is a long running series of these with larger sample sizes. Might cost a bit but you could probably come up with a reasonable hypothesis test and some solid results.
I'm starting to become tired of such liberal use of the words 'hack' and 'hacker'. _Maybe_ 'marketing hack' would have been ok, but honestly there isn't even any programming involved here.
On an entirely separate note, nice product name. And that does seem like an interesting method for picking one.
A hack does not have to be about programming. If you want to qualify hacks as 'marketing hacks' or whatever else, you would need to qualify your preferred kind of hacks as 'programming hacks' as well.
I debated the name and really tried not to, but I think it does fall under the definition of a hack... I could be wrong. Sorry if I mislead. Was not the intention.
