August 20, 2019
Leveraging CRM Data to Dictate SEM Investment
With automation a burgeoning trend in paid search advertising, Google Ads has been improving both the efficacy and number of smart bidding strategies it is offering. While relinquishing control of manual bid adjustments can be a hard sell for some PPC professionals, successful testing of these algorithms can result in additional time for deeper strategy initiatives. In addition to freeing up time, we have seen performance improvements of up to 200% through the proper use of smart bidding.
In this ultimate guide to automated bidding, we will explore the basics of our favorite strategies, how to properly test them, as well as tips and common pitfalls.
From enhanced CPC to target CPA, Google Ads smart bidding leverages machine learning algorithms and user-level signals to predict performance and set bids at the time of each auction. Using robust contextual signals such as time of day, location, device, remarketing list membership, the query itself, and more, Google predicts the likelihood of conversion for each auction and then sets the optimal bid accordingly. Google recognizes these signals in tandem, not isolation, to adjust for meaningful combinations of signals to positively impact conversion rate. In order for Google to do this successfully, there has to be sufficient conversion data to work with.
After significant agency-wide testing, we feel the following conversion and revenue-based strategies are most likely to yield the strongest performance: target CPA, target ROAS, and eCPC. For the purpose of this guide, we will focus on these options, though did want to briefly note all other available bidding options:
Target CPA bidding, formerly known as “conversion optimizer,” aims to capture as many conversions as possible within a specified CPA threshold. If you determine your business is profitable at a $20 CPA, you can set $20 as your threshold, and the algorithm will seek to maximize conversions within that. If you increase your CPA target, the algorithm will capture additional conversions. There is no conversion minimum to launch target CPA, though we have found stronger results with at least 15 conversions over the last 30 days — more is better, and will result in a shorter “learning” period.
The results below are a representative sample of what we’ve seen across a number of campaigns and accounts.
If you are optimizing toward a specific ROI goal, target ROAS bidding is a strong option to test. In this strategy, Google Ads will seek to maximize revenue within a specified ROAS constraint. A bit newer than target CPA, we are seeing more diverse results with target ROAS. We speculate this is because it is easier for the algorithm to predict a conversion than to predict the specific value of a conversion. Target ROAS also has stricter volume requirements and will not be available to all campaigns if there is not enough historical data from the past 30 days.
Enhanced cost-per-click (eCPC) bidding is a great option for those who want to introduce machine learning into their account but also maintain some degree of manual control. It is considered a semi-automated strategy, because some manual bid control is maintained. Enhanced CPC bidding leverages the same auction-time, contextual signals as other automated strategies to adjust the manual CPC bids you set. If the algorithm believes a user is likely to convert, it will increase bids to ensure your ad reaches that user.
While we have generally seen strong results with smart bidding, it hasn’t been a win in all cases. We always recommend testing these bidding strategies to see how they affect your account.
The Google Ads Experiments tool allows you to A/B test virtually any aspect of your account. Simply create a draft campaign, adjust the variable in question, and create an experiment, splitting traffic however you see fit between your experiment and control campaigns. The Experiments interface offers easy performance comparison reporting that indicates if statistical significance has been achieved for each metric. However, while the Experiments tool makes it easy to test different bid strategies against one another, it has one drawback: volume.
The more conversion data points you are able to provide the bid algorithms, the better success you can expect to have. By dividing your traffic and conversions in a 50/50 split test, you are giving the algorithm significantly less data to work with. This puts the algorithm at a disadvantage from the very start of the test, and the results of the experiment may not be fully indicative of how the bid strategy actually performs.
As a result, we recommend that you instead directly implement the bid algorithms in your campaigns and review period over period performance. Although period over period results make changes in seasonality or the competitive landscape more challenging to account for, you should be able to expect a higher probability of success for your test.
When launching your bid strategy test, there are a few items to keep in mind:
While not every test will be a win, by following these guidelines, the algorithms will have their best shot at driving results in your account. Even if performance is flat when you use manual bidding optimizations versus the algorithms, you should consider that a win for the bidding strategies. The time you save from using automated bidding can be reinvested into generating new creative ideas, growth initiatives, and other strategic efforts.
If you’ve made it this far, you are well on your way to mastering automated bidding. Before you can truly call yourself a smart bidding expert, however, there are some details you need to know. To complete our ultimate guide to smart bidding, we’ll outline some final tips and common pitfalls.
Portfolio bid strategies enable you to group multiple campaigns together into one overarching bid strategy. These can be used to provide the algorithm with additional data points when individual campaigns do not have adequate conversion volume to expect success or to run at all. Portfolio strategies can be set up in the Shared Library > Bid Strategies.
The bid strategy reporting interface is chock full of good information, including the status of your bid strategy (i.e. is it in the “learning” period still), how it has been performing and a history of changes to targets. This report can be found by clicking the link from your bid strategy column, and is only available for certain strategies like target CPA and target ROAS.
We recommend setting an initial ROAS or CPA target that is relatively close (within about 20%) to the current state of those campaigns. If your aspirational ROAS is a 5, but you’ve only ever achieved a 3, you can expect that trying to hit that 5 right off the bat will result in significant spend decreases and often lower overall performance.
Although we generally want to give the algorithms as much leeway as possible, we also feel setting reasonable CPC caps makes sense. This is to avoid extremely high CPCs on certain search terms. For example, if your CPA target is $50, an actual CPC of $50 would require a 100% conversion rate in order to meet your goal. While we have confidence in the algorithm’s ability, we feel predicting a 100% conversion rate is a tall order. Use your best judgement, but a CPC maximum limit of half of your CPA is a good general guideline.
While there is no hard and fast rule, you can expect less impressive results with target CPA bidding if you have fewer than 15 conversions per month in the campaign or group of campaigns you are testing within. For target ROAS, we recommend having at least double that number to maximize your chances for success. If you do not have adequate volume, using enhanced CPC bidding is your next best option.
It’s important that you do not use target ROAS or target CPA in budget constrained campaigns. This prevents the algorithm from working optimally. If you run into budget limited campaigns as you are testing these bid strategies, we recommend opening up your budget, or gradually increasing your ROAS target (or decreasing your CPA target) until you are no longer limited.
If you are leveraging back-end sales data to optimize your advertising dollars, it makes sense to use extra caution when using smart bidding. Often times, the lowest cost leads aren’t the highest quality, so there is the potential for target CPA to optimize away from quality. We recommend frequently reviewing the search query report and keeping a pulse check on quality metrics when you launch a new smart bidding test.
With your new smart bidding foundation and the above tips, we hope you are now more comfortable delving into the world of automated bidding. Between performance improvements and time savings, machine learning algorithms are the future of paid search and should be embraced as an opportunity.
For more digital marketing content, make sure to subscribe to our newsletter using the right-hand widget above. If you’re looking for more detailed help with campaign strategy, contact our team.