The cool but challenging thing about targeted advertising is dealing with all the moving parts.
Sometimes, for example, we may do our best to interpret an ad test, but in an environment where our analysis is handicapped by other mistakes that skew the test.
Let’s say our targeting is inadvertently too broad. An account manager has bid relatively high on a one-word broad match, using any of the following syntaxes:
There are a number of better-targeted keywords in the ad group. They generally perform well. The very broad match has its uses, but if it’s bid as high as it is (let’s say 90 cents when 40 cents might be more suitable for our purposes), it’s generating a lot of dissatisfied searchers along with a small number of reasonably targeted prospects.
In this environment, the “winning” ad is too tied to how good that ad is in dissuading tire-kickers from clicking. In short, that ad test is really about making the best of a bad situation.
Well, whoever created that situation can fix that situation. If you’re doing a job and the ladder is on the wrong side of the building, it doesn’t matter how hard you work. Nothing gets done until you move the ladder.
What if we brought the targeting into a more conventional range? Where the majority of the impressions were served to reasonably well-qualified prospects?
Now, you’d be truly testing which ad was best under more favorable conditions. You’d be optimizing your results rather than just managing mediocrity. And even further refinements could take place within a lab that provides appropriate conditions for testing.
The same might be true if you had failed to exclude a demographic group that can’t buy your product (or who very rarely does) – say the 18-24 age group – or if you failed to manage gender demographics to your KPI’s. In those situations, where a good chunk of those seeing that ad are by definition non-buyers, it doesn’t matter if they click or not (or more precisely, you’re in the position of hoping a lot of people don’t click, instead of ensuring that they don’t see the ad in the first place).
Consider that many efforts at analysis in a “broken” ad group will be mirages. Improve the targeting so that your testing can proceed in a more optimal environment for interpreting test results. “Winning” results, ideally, are about thriving – refining a very good swath of ad targeting to a great one – not mitigating a fundamentally flawed setup.Read Part 25: Five Ways to Keep Smart Bidding From Acting Not-So-Smart