Ad copy testing is one of the foundation blocks of PPC campaign optimization. Yet, as any PPC manager knows, it can be laborious, time-intensive, money-consuming, or take eons to get enough data to make a call and move onto the next test.
Jonathan Nelson and Yuanyi Zhang developed and recently launched AdProof, a new “user testing” platform for running A/B ad copy tests, which harnesses crowdsourcing to help make ad copy testing faster and cheaper. Tests cost as little as $1.00 to receive results within one day. An extra $5.00 includes user feedback.
Experiments start with a keyword. The system will show you the estimated number of impressions, cost-per-click and number of other advertisers competing for that keyword, and then auto-generates two test ads from the keyword you enter. You can make whatever edits to the ads you want, or use your own original copy. Choose your network (Google AdWords or Bing Ads), how quickly you want results back, and if you want user feedback, then hit go.
Ad placements are evenly rotated according to the predefined position — you can choose to run the test in the top or right-rail positions. The experiment results page shows you the average position of the ad during the experiment.
The system scales the number of testers for each experiment by crowdsourcing users, about 80 percent of whom are English speakers in the U.S. “This allows us to keep our pricing extremely cheap and provide fast results,” says Nelson. As far as total clicks per experiment, that also depends. Currently, the click algorithm closely follows the usability research from Jacob Nielsen, which holds that small testing numbers make for faster and better tests. However, the company says, in the near future, the total testers or “clicks” will be controlled 100% by the customer.
When I asked about statistical relevance and accuracy, Nelson replied, “We eat our own dog food here, and the user testing data is extremely accurate. Not to mention user feedback in the form of comments, which is something that was never available until now.”
When an advertiser launches an experiment, testers are prompted with a window that describes the keyword search and testing scenario. For example, the keyword might be [roofing tiles] and with a scenario of “You are looking to buy roofing tiles.” After the tester accepts the experiment, they are presented with either Google or Bing search results for the keyword query. The advertiser’s A and B variations are evenly distributed along with other relevant PPC ads for the keyword query. “This allows us to execute the experiment in the most natural way possible,” says Nelson.