Google Ads bidding algorithms are getting old. Since their inception in 2016, they have gone through countless iterations through multiple campaign types. However, the core concept of each algorithm has mostly remained constant. Over time, the campaign types have been the major driver of change that has slowly pulled more and more campaign settings into the black box of Google’s machine learning, especially in the Shopping arena.
First came Smart Shopping, which effectively required Shopping campaigns to use a Smart Bidding option that relied on data algorithms in order to work at full capability. Then came Performance Max campaigns, which kept most of the “under-the-hood” information out of the advertiser’s reach.
But we still have controls to play with if you know where to look. There are three levers buried in the settings of Google Ads that allow advertisers to maintain control of machine learning. There is a confusing wrinkle in the settings that manipulate algorithmic learning: they are not actually in the campaigns at all.
Under Tools and Settings in the top right navigation bar, there is a drop-down in the Shared Library called Bid Strategies. At first glance, this window does not have that much valuable data, but in the left sidebar, you will find Advanced Controls. From here, you can inform all the algorithms in your account about upcoming market shifts (Seasonality Adjustments) or how to exclude data after the fact (Data Exclusions).
Seasonality Adjustments
Seasonality of industry is one of those things the algorithms are supposed to anticipate, but often we see it take too much time to catch up to shifting market conditions. Generally, holiday trends take place over the entirety of Q4, which gives the algorithm time to adapt. But for more immediate shifts, like a Black Friday sale too good to miss or a two-day spike in sales because you were featured on a TV show, Google’s machine learning just doesn’t have the ability to shift in time. This is generally a good thing, as it maintains the stability of the algorithm, but sometimes we need to take control of the wheel. With Seasonality Adjustments, you can tell the system you expect your conversion rates to have dramatic changes for a short period (best for one to seven days) and to not “overreact” to that data.
Data Exclusions
Another handy tool found just next to Seasonality Adjustments in that Advanced Controls area is a tab called Data Exclusions. Look, conversion tracking isn’t perfect. There’s a lot of complexity just in linking your Google Ads account to your website, let alone your internal CRM or ecommerce platform. When conversion tracking isn’t functioning correctly, the algorithm starts working with incomplete or inaccurate data. This can create a feedback loop where the algorithm starts to break down, and this can spiral into a completely broken algorithm that needs to be reset.
The solution to this is Data Exclusions. If you can pinpoint exactly when conversion tracking went haywire, you can tell the algorithms to ignore that time frame from this tab. However, this function only works well with shorter time frames. Anything longer than two weeks will be largely ineffective.
Data Exclusions also works for things like large sales that can be attributed to channels outside Google Ads that need to be excluded from performance data.
Brand Exclusions
Our third way to control algorithmic spending, especially on Performance Max campaigns, is to use brand exclusion lists. This function allows you to exclude your brand name from your searches so as to not cannibalize from your Search campaigns. You’ll just need to go into the Performance Max campaign settings, and at the bottom, under Advanced Settings, you’ll have the option to exclude the brands on your website from the campaign targeting.
If your brand or your competitor’s brands aren’t there, they can be requested to be added.
As machine learning continues to change our day-to-day workings in Google Ads, our ability to manipulate algorithms is sure to shift. Google will hopefully add new ways for us to give more data to the machine learning to improve results, especially as more campaign types evolve.
