Google Buys Arizona and Southern Idaho

Hypothesis: If I use a shocking blog post title, I will receive more traffic than any previous blog post on this site.

Um, so Google didn’t start buying U.S. States? No, not yet. My apologies; this is a post about the importance of having a hypothesis before optimizing search campaigns. Thanks for participating.

In our data-centric, direct response world of Search Marketing (PPC+SEO), you can basically run an experiment on everything: different ad copy, bids/positions, landing pages, match types, day-parts, geo-segmentation, meta titles, keyword density, calls to action, and much, much more. Because we have this ability, does that mean we should test just to test? Of course not. Does that mean we should always be testing something? Probably.

With all of the available options, one should choose an experiment based on its potential for positive impact on campaign goals (aka, KPI’s). From looking at digital media ROI reports or on-site behavior in web analytics, there is a vast amount of data to comb over and ask, “Is this what I expected?”, “Can performance be improved?”, and “…by how much?” Further, compare the data to industry benchmarks to see if your expectations and performance goals are in line. Speculate on why things are happening and what influences them.

Once you’ve identified a testing opportunity, step one is to write a hypothesis. In other words, write down the steps you are planning to take and the improvement you expect. “Copy A will outperform Copy B in Click Volume, CTR, and Conversion Rate by using ‘Official Site’ in the ad title.” “The Category X landing page would have a higher entry to conversion rate by using product images over people images.” “Broad matching keyword phrases containing over three words, I’ll gain more high converting ‘tail’ traffic.” Wrong assumptions will be made; it’s OK, as not trying is much worse than failing. The next steps of a good experiment include establishing baseline data and KPI success metrics as well as determining the test duration and projecting performance. Remember, testing always starts with a hypothesis. Don’t just test to test, have a goal and design that test around it.

Posted by: Jeff Campbell, VP Product Development & Innovation