When I talk to customers, I use a phrase like, "We can, but we have to make sure that we lubricate the wheels of the machine first." My industrial age jargon actually means that we need to make sure that we don't force the algorithms to relearn in ways that adversely affect our performance. I'll spare you arguments as to why you should turn to machine learning in your digital advertising work if you don't already know – lots of people out there can help you. If you're struggling with automation, check out Lauren Rosner's Abandoned Control blog or How I Learned How to Stop Worrying and Love PPC Automation.
Below are the most useful elements to understand how algorithms work in your PPC campaigns.
The target ROAS as an intelligent bid strategy is best suited for campaigns that are NOT budget constrained and have at least been successful 15 conversions in the last 30 days. The strategy achieves the highest possible conversion value with the same return on advertising expenditure. This strategy doesn't work well in a hyper-cemented account structure. The target CPA works very similarly, except that it is optimized for a CPA value and does not have a minimum conversion volume.
When using target strategies, it is important to understand what you are asking of the strategy:
If you have an aggressive CPA or ROAS goal for your industry, such as 20x ROAS, ask the algorithm for it limit your risk By severely limiting the number of auctions you want to participate in. They're effectively saying, don't bid on auction unless you're damn sure you can get me a 20x return. This leads to the fact that the algorithm becomes very selective and participates in auctions very carefully. This can lead to a decrease in the percentage of impressions, an increase in CPCs, and an overall decrease in volume. Ironically, automation takes a lot of data to make decisions to work optimally. Hence, it is important to be REALISTIC in setting your goals. Here are some rules to keep in mind:
- The first time you switch to one of these automated bid strategies. Set the initial goals based on yours last 30 days CPA or ROAS.
- It takes 7 to 14 days for the algorithm to be recalibrated. So give it time to get set up and not make any changes. It only begins with the "performance" when the learning phase is over.
- Move your CPA or ROAS goals in increments of 15% to 20% at a time. For example, if your ROAS was 5x for the past 30 days and you want to get closer to a 10x goal, switch to 6x first – let it run for a few weeks. Then move 20% again. Avoid aggressive shifts when doing CPA or ROAS on the boat not rocking.
- Feed your algorithm by overlaying the target audiences. Set this to "observation" and make the appropriate bid adjustments.
If you're struggling to scale campaign volume, your goals can be the limiting factor. Try increasing your CPA or decreasing your ROAS goals in increments of 15 to 20%. This does not necessarily result in poor performance. By lowering your bar, you are effectively telling the algorithm that you are not that risk averse and ready to take risks as you try to achieve your goal. If your accounts are seasonal, I highly recommend lowering your ROAS and increasing your CPAs at least 1 week in advance to allow the account to scale during times of high seasonality.
Maximize conversion value or maximize conversions are budget based strategies and should be used for campaigns that are regularly limited by budget or that spend most of their daily budget. The minimum criterion for these strategies is 20 clicks / day. In order for max conversions to work effectively, make sure you only track important conversion actions or set up conversion action sets (linked to the PPC hero article). These strategies work best for broad match type keywords, which will allow more traffic and the strategy will provide more data to learn.
When asking for the max, ask the strategy of ending the auction and capturing as many valuable conversions as possible based on the budget provided. It has no impact on your other campaign inputs and does what it wants. For example, when participating in auctions, these strategies DO NOT consider the following:
- Adjustments to the device offer
- Adjustments to the location requirement
- Ad schedule
- Old age
- HH income
These strategies don't care about CPCs either. If you have keywords that have CPC goals set, move them to another campaign. This is not the strategy for these terms. A common misconception when using these strategies is that you have to start with eCPC and then move up to Max Conversions. You do NOT have to do this. If you do this, the algorithm will have to learn and recalibrate twice. You can start new campaigns using the "maximize conversion" or "maximize conversion value" strategies.
If you're not sure if a bid strategy is working for a particular campaign, see the bid strategy report. This can be found here:
Hover over the bid strategy type in the Bid Strategy Type column and click:
The Bid Strategy Report looks something like this and contains information to help you optimize your strategy. For example, actual CPAs have been going up lately, so I might want to adjust my goal to make sure I can capture more volume.
Even if it is difficult, you need to plan periods of study for your intelligent bid strategies. The general recommendation is 7 to 14 days with no changes before you notice significant performance shifts. One of the biggest dangers for marketers is to preventively label a strategy as ineffective before the end of its learning curve. Plan ahead, make sure expectations are understood ahead of time, and when in doubt, consult your bid strategy report to learn how to adjust.