I advertise a fair bit on google, as well as other sites. I was getting annoyed with the system recently, carrying out multiple changes to my Gratuitous Space Battles campaign, and trying to tweak things to avoid having my ads shown on flash-games compilation sites that look horrid, and trying harder to get them on QuarterToThree, RockPaperShotgun, etc.
Then I say back and thought about all my tracking systems, and a book I read called supercrunchers, and zynga, and I came to a different conclusion about how to manage it all, a conclusion I am now testing out.
I have a system, whereby I know if a visitor from a google advert click ever hits the GSB demo download link, or the buy page. I consider these to be pretty good indicators of purchase intent, and what I do is to equate them to a monetary value as a conversion. I derive this basically by dividing up the income from the sales by the extent to which those events happen. It’s crude, but not *that* crude.
That means I can say (for example) that a demo download is worth $0.40. That means if I can get enough clicks from enough sites on google to generate that download, for less than $0.40, I’m winning. This is especially true given the potential long term lifetime earnings from a new customer.
Anyway, google can already automatically handle all this crap for me. My analytics can crunch out the top 25 ROI websites for investment, over the last 3 or 4 months of advertising. These wont be the cheapest, or the most suited to my game, but the ones where the amount of clickers who actually go on to download or consider buying is the best value for money.
So as an experiment, I ditched every single site I advertise on through google, all of the hand-picked ones, like RPS, and various space games and sci-fi sites, and replaced them purely with the 25 top ROI sites for those ads, as judged by the google stats. In other words, I am binning my pre-conceptions and guesswork and opinions and ideas, and purely trusting the data. Because conversions take time to register, I reckon it will be a good 5 days before I know if this simple, and entirely data-driven process is working. If it is, I may never change it. Why would I?
So my new plan for ads is two phases:
phase I: let this run another 4 days at elast, without changing ANYTHING, and observe the data.
phase II: if the data suggests I am making money, double the budget. In a month, if the data still holds, quadruple the budget.
I’ll be sure to blog the outcome :D