September 12, 2019
A/B Testing in Google Shopping to Prepare for the Q4 Holiday Season
You remember the fashion fads from the past several decades, right? Shoulder pads and excessive hair in the 80s, grunge or gothic versus a classic vintage look in the 90s, to the frosted hair spikes of the 00s (thanks Justin Timberlake and Guy Fieri).
Well, the digital marketing industry has had its fair share of fads, as well. Way back in the day, broad match was everywhere, and negative keywords were the most important tool. Standard Text Ads turned into Expanded Text Ads, and now they’re even going out of style thanks to Responsive Search Ads. Automated bidding was in, then out, and now has come back with a vengeance.
But what’s the newest digital marketing fad that should be in the discussion to add to your strategy? Smart Shopping.
What Exactly Is Smart Shopping?
For starters, Google views Smart Shopping as entirely separate from regular Product Listing Ads (just another acronym that’s been retired), or manual shopping. In fact, Smart Shopping is viewed as so separate that Google has even dedicated a team exclusively tasked with Smart Shopping’s growth, maintenance, and of course performance.
How Are Smart Shopping and Manual Shopping Different?
One explicit distinction between Smart Shopping and Manual Shopping is their inventory makeup. While Manual Shopping ads live on the typical SERP and Shopping tab, Smart Shopping can also showcase product ads across the entire Google Display Network, Gmail, and YouTube as well as utilizing remarketing audiences. Because of this, Smart Shopping ads must include both text and image assets (similar to Responsive Display Ads).
It’s not all good news though. There are some major drawbacks to Smart Shopping and almost all of them are related to the lack of information Google makes available to advertisers. There are no audience insights or search query reports.
However, the characteristic that differentiates Smart Shopping the most is that the bidding strategy utilized is not algorithmic, but rather machine learning. Algorithm-based bidding strategies take all of the data points at its disposal at any given time to produce a bid for a given auction based on the target goal, whereas machine learning does all of that and also incorporates the results of past auctions as additional data points. Long story short, Smart Shopping should continue to improve performance over time.
Why Should You Test It?
From a performance standpoint, Metric Theory clients on average have seen 35% revenue growth and 15% improvements to ROI as a result of utilizing Smart Shopping.
If that’s not enough, then maybe the time savings will be. It’s not just a bidding algorithm – there also aren’t negative keywords or bid modifiers (Hour-of-day, Day-of-week, Geographic, Device, etc.) to deal with.
Even if you are currently experiencing strong shopping performance, that is not a reason to not test smart shopping in some capacity.
Why Shouldn’t You Test It?
The reasons for not testing are more a matter of business goals than anything else. If segmenting between first time and returning purchasers is a high priority for the account, then Smart Shopping currently doesn’t offer a way to continue that strategy.
Accounts with heavy call volume that are not being accurately tracked in ads also presents some difficulty as the bidding strategy doesn’t have access to the key data points it needs.
Also with no search query report to sift through and add negative keywords from, it may not be a good idea to test Smart Shopping if it’s known that the feed being utilized is bare bones or has structure that hasn’t been reviewed for best practices.
How To Test
There are three main ideologies we’ve seen utilized most often when testing Smart Shopping.
Option A: Identify your poorest performing product group/s and test those exclusively within a Smart Shopping campaign. If there are product group coverage overlaps between your Manual Shopping and Smart Shopping campaigns, Smart Shopping will take priority and take up the overwhelming majority of quickly within just a few hours, so there’s no need to exclude the product groups you want to test from Manual Shopping.
Option B: A recent addition to Smart Shopping that Google has rolled out is the ability for strong location targeting. This comes in handy as you can A/B test Smart Shopping against your current Manual Shopping structure while also being able to limit the overall performance impact via state or country geographic restrictions.
Option C: Of course you could always make a clean switch from Manual to Smart Shopping if performance necessitates or there are few factors that would influence a period over period analysis.
One major thing to keep in mind when analyzing Smart Shopping performance is that since its inventory also includes Google Display Network, Gmail, and YouTube via remarketing, you should be incorporating performance from your Display Remarketing campaigns to evaluate performance more effectively.
We’ve been testing Smart Shopping for clients across a variety of industries, all with different reasonings behind the test, and have seen overwhelming positive results showcased below:
Automation is the present and future, and as digital marketers we can’t hesitate to embrace the most current technologies that are at our disposal. While the initial results that we’ve seen in Smart Shopping across a range of accounts have been very strong, it’s important to test before rolling out immediately and ensure that both Remarketing and Shopping performance are being taken into consideration when evaluating success.