Google Smart Shopping Campaigns: The Power of Machine Learning in Paid Search

July 30, 2019

Google has made a concerted effort to adjust the shopping experience for both consumers and advertisers over the last year.

Shopping has become a major area of emphasis as Google tries to compete with its top competitor in the space, and one that has historically dominated online ecommerce: Amazon. Google Express was one of several 2018 initiatives designed to help Google compete with Amazon from the consumer experience side. From the advertiser’s side, Google has always been ahead of Amazon. A more user-friendly UI experience, easy setup, and a myriad of targeting options have all been staples of Google’s advertising platform. The most recent change to shopping from an advertiser’s perspective is the introduction of Smart Shopping.

Introducing Smart Shopping

At the Google Marketing Live event in July 2018, Google announced that it rebranded its machine learning shopping campaigns from Universal Shopping ads to “Smart Shopping” campaigns. The fundamental idea behind Smart Shopping is to give Google all the control: You tell Google what you want to reach as a goal, and Google will do the rest. Unsurprisingly, this left the advertising community feeling skeptical. As marketers, one of the best aspects of search engine marketing is the complete control over the user experience and advertiser goals. The ability to manipulate products that are shown through bids, negative keywords, promotional remarketing, etc., were all levers we could use – so the thought of giving that up was intimidating and nerve-racking, to say the least. However, after shopping campaigns began to slow down for one of our largest ecommerce clients, and one of the largest companies in its vertical, we decided to give Smart Shopping a chance. The results speak for themselves.

Machine Learning in Action: Taking Smart Shopping for a Test Drive

Our first aspect goal was to regain efficiency. Over the last year, the space had become extremely competitive and CPCs were on the rise. Even though conversion rates were still higher than they had been, the CPC increases were too high to overcome and ROAS began to decrease. The decline reached a point where they were no longer profitable campaigns. So, the first step was to regain that profitability. We set up the campaigns to test a small segment of the feed, choosing an area that would have high enough volume to get actionable data, yet low enough that it wouldn’t send the account into a tailspin if the test didn’t produce the desired results.

In order to start Smart Shopping, we needed logos, text ads and an audience of at least 100 people in a rolling 30-day window. These are Google’s requirements, as your ads will show up on normal shopping SERPs, GDN, YouTube and Gmail, with no way of opting out of anything. Once that was situated, we launched. The system automatically prefers Smart Shopping campaigns over your normal shopping campaigns, so impressions immediately shifted over before I could even pause the original shopping campaign.

The initial data was great. Efficiency increased to nearly double of what it had been. This meant that we could now look towards maximizing revenue as best as the system could. Volume nearly doubled within the first week, with enough data to quickly call the Smart Shopping campaign a winner with statistical significance. We slowly started to roll it out to other areas of the feed, and within three weeks, we had adjusted all the shopping campaigns to Smart Shopping. After running it for several months, we are pleased with performance from an efficiency and revenue standpoint.

ForwardPMX’s Paid Search team began moving advertisers over to Smart Shopping at the start of 2019. The advertisers we have moved over to Smart Shopping have all had success at some level, whether efficiency or volume. Volume has been up between 70 to 125 percent, depending on advertiser and product type. And while that’s amazing, it still isn’t the best part. The most impressive highlight is that we were able to gain that much revenue while increasing ROAS. Efficiency gains are up between 25 to 55 percent, in some cases increasing net profits by 400 percent.

Getting Started with Smart Shopping: Key Takeaways

There are a few key takeaways on how to best set up Smart Shopping campaigns and strategies to maximize efficiency and revenue:

  1. First, group items by similar ROAS/CR. This will allow the system to push in the areas that are working, without sacrificing much in terms of revenue.
  2. Second, start with ‘maximize conversions’ as the main goal to help build volume.
  3. Once volume is established, move to a ROAS model. Default ROAS on these campaigns are $2.00 USD, so if that is too low, there’s plenty of room to increase revenue by moving to a ROAS bid strategy. If $2.00 is too low, then definitely adjust the targets in a ROAS strategy to help you maximize the revenue under your higher ROAS goals.

As a lifelong skeptic of Google’s intentions with advertisers’ money and users’ data, giving Google and its machine learning technology relative autonomy and the ability to drive our campaigns was downright scary. In this case, I think its machine learning has gotten it right, as many advertisers have seen similar success to what we’ve experienced with Smart Shopping. I’m not running to the soapbox just yet, though I am quite pleased with the results of this feature and would recommend Smart Shopping to any advertiser that uses shopping campaigns.

How do you feel about applying machine learning to your paid search campaigns? Share your experience with Smart Shopping in a comment below!

Follow ForwardPMX

Our Newsletter

Sign up to receive our monthly insights.

  • By submitting your Email Address, you are agreeing to all conditions of our Privacy Policy
  • This field is for validation purposes and should be left unchanged.

You May Find These Interesting

Our Visit to the Forest For Change

As humans, we’ve historically loved to dominate and take ownership of things. We saw something, free-formed, colored outside the lines, and said, “Nope, I didn’t say that was allowed.” So, it makes sense then, that back in the 18th century a man named William Chambers...

read more