Skip to content

Machine learning in advertising: think about it

Anyone running their own search advertising campaigns will at least have been looking at Google’s new ‘Performance Max‘ campaign strategy for a while. This is the one where Google’s machine learning technology largely runs the campaigns. As advertisers, we upload ‘assets’ including images and video, and let Performance Max try to get conversions across Google’s range of channels including Search, Display, Discover, Maps, Gmail, and YouTube.

This can be seen as the lazy way to use Google Ads, and it’s not unreasonable to dismiss it if we want complete control over what we’re doing. However, it can still be a potentially interesting tool for those of us who like to take a more hands-on approach. At BMON we’re dipping our toe into the water with some clients, notably those with obvious and measurable goals such as enquiry form completions and online stores.

No surprise

It won’t be a surprise to discover that like any machine learning technology, Performance Max has to experiment wildly to home in on what works. This can be expensive if not managed carefully. It can also cannibalise existing campaigns, in particular the low-hanging fruit such as brand-focused ones, so we need to stay on top of what we allow it to do. According to Google, where there’s an overlap, a standard search campaign gets priority, but if there’s a small difference, the campaign or ad with the highest Ad Rank gets used.

The key to a successful machine-managed campaign seems to be to ensure the goals are specific, measurable actions. This isn’t always possible for many B2B companies, whose priority is simply to generate solid, engaged website traffic at a reasonable cost. However, there may be a place for Performance Max in many advertisers’ overall advertising efforts. Anyone wanting to investigate further should have a read of last month’s article 11 Tips To Get The Best Out Of Performance Max Campaigns on Search Engine Journal.