One of the keys to creating successful pay-per-click search ads has always been the practice of “A/B testing.” This method of online ad management involves running two ads simultaneously (essentially rotating between them) for a period of time, then keeping the ad that performs better and replacing the other. By doing this continuously, you hone your ad messages until they resonate with the target audience.

It’s a very sound scientific principle and as our resident pay-per-click expert, I can tell you that it works. But not everyone uses it. Some people simply create ads and let them run on autopilot, settling for whatever performance they might deliver.

This is their choice, of course. Or at least it used to be. Recently, Google began to experiment with a system that A/B tests on your behalf, without asking you or getting your input, whether you like it or not – unless you delete their ads manually. Did I mention that they are doing this without asking their customers?

I saw this firsthand as I was doing some maintenance on a client’s Adwords account recently. I was stunned to see several ads in the rotation that I did not create (instead they said “Added By Adwords.”). Google’s omniscient algorithm (“all hail the algorithm”) had collected information from the client’s website and automatically created a number of ads to mix into their ad groups. Fortunately, none of them were bad or wrong – they just weren’t good. Worse, Google’s algorithm (“praise the algorithm”) didn’t understand some of the subtleties of the client’s products – or more critically – the consumers who used the products.

Way back in the beginning of online search advertising, we put web programmers in charge of pay-per-click marketing. It only took a short time to learn of our mistake. You see, Google had already done all the programming, so it wasn’t about code at all – it was about crafting messages that effectively attracted clicks. That doesn’t call for a background in computer languages, it calls for a deep understanding of the target audience and an ability to assemble meaningful words (and sometimes images) to engage them.

In other words, it takes a creative department to make the most of Google Adwords. It simply cannot be done automatically via an algorithm (“no offense, oh great and wonderful algorithm”). There’s just too much context involved.

The ads that Google wanted to insert were simply sentences cribbed from our client’s web content, rearranged to fit a specific format. There was no thought or strategy put into them, and it showed. Worse they showed a lack of understanding of how A/B testing on Google Adwords works best:

(1) You start with what is working so far and make subtle changes to improve your ads. If you start from scratch each time, you tend to miss more than you hit.

(2) Ironically, once you have a successful ad, you may not want to A/B test as aggressively. Think about it – if you’ve honed your message until it is effective, why dilute your results with an ad that doesn’t work as well. I presented this theory to one of the experts at Google via a phone call a few weeks ago, and it was like I had explained the meaning of life to him.

“You’re right,” he said slowly as it sank in. “Your click-through rates are higher than I have ever seen. Why would you experiment with those?” In other words, if you have an ad that is a finely tuned winner, why dilute its performance by testing an unproven ad that ends up working half as well? The “B” ad’s lame performance drags down the metrics for the “A” ad.

That’s not to say you can rest on your laurels. Just the opposite – the goal is always to create winning ads. But once you have created a winning ad, you need to get the most out of it. And that means trying to improve it slowly and in a way that doesn’t undermine its success.

Underperforming or average ads? Those you need to A/B test much more, obviously.

Google’s algorithm (“forgive me, all-powerful algorithm”) can’t see these subtle differentiations, at least not yet. It takes a more binary, shove-em-down-your-throat approach to the scientific method.

Thankfully, I caught these “helpful” ads before they watered down a couple of very successful Adwords campaigns. Our clients have chosen to take pay-per-click seriously by paying for us to manage their accounts, and it pays off. It’s strange, but one of the most important jobs that an online ad manager does is to stop companies like Google from screwing up their own systems.

I’ll keep an eye on the algorithm (“hey algorithm – stay out of my stuff”), and I suggest that you do, too. Sometimes Google doesn’t know when enough is enough, and we need to remind them that they serve the internet and not the other way around.