Not everything Google told me is true

The buzzword in AdWords these days is “machine learning”.  The algorithms are going to learn your advertising and optimise it better than you can.

There has been a flood of articles from Google about how wonderful it is, and a flood of comment in the discussion groups about how bad it is.  Let’s assume that the truth lies somewhere in the middle, but that still doesn’t tell us when to use it and when not.  Worse AdWords users are actively prompted by Google to adopt machine learning any time a campaign reaches the minimum requirements.

My pet theory is how retail (webshops, physical shops) works differently in AdWords than other businesses.  Promoting a software package or a plumbing business is one thing, but a webshop has hundreds or thousands of products, and most of those products don’t get enough clicks for machine learning to figure them out.

My answer at the moment is to run dual campaigns.  I let the machine learning handle the high runners while I work manually on the rest.  A smart “filter” transfers products to the automated campaign exactly when they can best benefit, so we can keep growing.

Nevertheless, this new technology is a moving target.  A lot of clever people are busy working out new and inventive ways to use it.  Who knows where it will all lead.

Postscript

Since the early days of Google’s advertising, experienced users have referred to what they call the Google Stupidity Tax.  It’s the wasted money new advertisers pay for using Google’s default settings.  So nowadays I wonder what happens when you get two automated accounts bidding against each other.  The automatic systems just keep pushing the price up?  

As they say about casinos, “The house always wins”.