Craft a smarter ad bidding strategy with machine learning in 3 steps
Step 1: Choose the right bid strategy
Set up conversion tracking and then think about what you’re trying to achieve. Is it increasing conversions, decreasing cost per acquisition (CPA), getting a specific return on ad spend (ROAS), or something else? Once you’ve defined your goal, pick the right bidding strategy to reach it. There are a variety of Smart Bidding strategies you can choose from. Maximize Conversions, for example, lets you get as many conversions as possible within a certain budget. But let’s explore how two companies used two other strategies.
If you don’t have budget constraints, you might consider the Target CPA strategy. Skechers, for example, faced an increasingly competitive footwear environment in Spain and needed to not only boost sales but raise brand recognition. The Target CPA strategy let the team define a bid amount that would achieve the highest number of conversions at a specific CPA. According to Skechers, this increased their conversions by 214%.
Target ROAS could be the right approach if you have a campaign that’s already driven a certain number of conversions, and you want a specific ROAS. Swedish clothing company Happy Socks needed to continue its rapid international growth. That meant scaling its advertising performance in each market, without having to invest a lot of time and effort. It used Target ROAS bidding to optimize all its search campaigns toward one ROAS goal, allowing the company to increase sales while still meeting its profitability goals. Happy Socks reported a 30% increase in ROAS across over 20 markets, a 20% cost reduction, and a 10% increase in sales as a result.
Step 2: Wait
Patience is a virtue when it comes to machine learning. After you’ve implemented your automated bidding strategy, make sure you allow for a proper waiting period before running a performance analysis.
Smart Bidding algorithms typically need a learning period of one week. But how long you wait really depends on how much conversion data is available, as well as the conversion delay, or the time between click and conversion. Check your bid strategy report to see how many days are left in your learning period. During this time, don’t make too many changes to your campaign.
Once your learning period is over, let your campaign run for a few weeks more. Then figure out the standard lag time for conversions (the average time it takes for a click to result in an online conversion). That’s how long you need to wait to run your analysis.
For example, let’s say your standard lag time is three days. Week one of your campaign could be your learning period. Weeks two and three are when you let your campaign run. Three days after that, you can start running performance analysis on Weeks two and three.
Step 3: Keep optimizing
You’ll free up a lot of time by no longer doing manual bidding. Use that time to get more strategic about optimization. For example, you might tweak your ad creative, improve your landing page, or design a better mobile shopping experience on your retail site. All these optimizations will help an automated bidding algorithm perform even better.
By choosing the right bid strategy, giving enough of a waiting period before analyzing your performance, and continuing to optimize, you can automate your bidding and make machine learning work for you.