Using negatives to shape traffic

Google, it can be guaranteed, will match some shoppers searching for “buy dozen red roses” to the Red Roses ad group. That’s because its machine learning doesn’t care (or know about) your frail human assumptions, and it wants to see what works and what doesn’t work over time so that it can put together its own model of what to do.

This is where human can still outperform machine. You’ve written your ad in the Dozen Roses ad group to do what a great ad should do, (be relevant, prequalify, and entice) for searches around “dozen roses”, you’ve set your landing page to where you sell roses by the dozens, and you’ve bid on the terms (or set automated bidding) according to your expected returns.

As brilliant as Google’s AI is, and as convincing the arguments are to get out of its way, you’re probably right to have to have your way and try to shape the “dozen roses” traffic to the ad group you want it to go to.

By adding “dozen roses” as a negative to the Red Roses ad group, you’ll prevent Google from getting too experimental for “dozen roses” traffic going to it.