The Undo Button

Richard Fergie writing on E-Analytica illustrates perils for marketers when trying to outsmart ad-tech algorithms:

Before Christmas I ran a brand bidding incremental value test for one of my clients. The results showed that nearly all PPC brand traffic was canibalized from organic. We paused the brand campaigns. This is about what happened next.

Mid January, competitor ads on the brand name terms became much more aggressive. We uhmmed and ahhed for a bit because we didn't want to waste money with no direct return but it was also obvious that these competitor ads were causing brand damage. At the end of January the decision was made to re-activate the brand campaigns.

"The undo button doesn't quite work the way you think it does."

For a variety of reasons (very generic brand name, high rolling competitors, other Google subterfuge) this account had always had a high CPC for brand terms (over $1). When we turned the brand adverts on again the CPCs for the brand campaign were the highest in the whole account. This was not good as it was eating our very limited budget.

What happened here is instructive. He had made a decision to run ads against the brand's words (this is the equivalent of running an ad linked to Amazon for people who search for the word "Amazon" in Google), and though it was nominally successful it was apparent that nearly everyone clicking on the ad (and costing money) was someone who would have clicked on the regular organic search results anyways.

So you start out with people Googling your brand name and clicking through to your site for free, and and up with people Googling your brand name and you paying a dollar or more for the same click. Whoops. 

The problem came in three stages. First he noticed his competitors were also bidding for the same words, and driving his costs up, and then when he decided to stop paying for those increasingly expensive clicks, he ended up with less traffic than when he started the whole experiment, and was forced to return and pay even higher prices. Spending money and effort to get fewer visitors is not a good way to make an advertising client satisfied.

Thankfully the story has a happy ending:

To figure out what was going on here I made the following assumptions:

  1. For these queries and advertisers, quality score is essentially click through rate.
  2. Google's estimate for our CTR remained unchanged from when the ads last ran.
  3. The estimate for the CTR of the competitor ads has increased since we stopped running brand ads.

This meant that our competitor's ad rank improved because we left the auction. When we returned our CPCs were higher because the actual CPC is the ad rank of the advertiser below divided by your own quality score.

Using this reasoning I predicted that brand CPC would fall as our competitor's CTR reduced now that our ads were back at the head of the auction.

Two weeks later it looks like I was right: the brand CPC is back to where it was.

An important lesson that with a flawed plan, an online marketing campaign is actually capable of causing real harm, beyond just being ineffective or wasteful. 

Leave a Reply

Your email address will not be published.