With the increase in automation offered by Google Ads and Microsoft Ads (whether advertiser-appreciated or not), one would be wise to not simply accept automation and utilize it but to optimize it, as well.
Wait, how does one optimize automation?
Doesn’t one simply push the button to move to Target ROAS or Smart Shopping campaigns, then sit back and crack open a cold one while the machine does the heavy lifting?
Automation isn’t fire-and-forget.
In fact, sometimes your job as the babysitter may be more hectic than you would have expected as you troubleshoot and guide the machine learning algorithms.
In this article, we’ll take a look at the bigger picture of managing automation and analyze the theory behind making decisions, so that the tactical process is better guided.
You’ll find tips for building the right framework, which can then be used for making solid automation decisions.
When it comes to making automation optimization decisions, we increasingly use something called the “Magic Wheel of Automation” as pictured below, in this state-of-the-art and expensive (most likely worth millions in NFT) chart.
While many people think automation is something to be set and ignored, aside from a small adjustment every once in a while, I encourage you to think beyond that.
When it comes to the way automation works, it is important to realize that machine learning (ML) is based on the data inputs it is given.
This means if the data inputs are incorrect, false, unclear, or incomplete, the ML algorithm will likely miss the target, no matter how intelligent that particular algorithm happens to be.
Once automation doesn’t appear to work, we should obviously investigate our data inputs to ensure they are complete (is your conversion tracking code placed correctly)?
But I have found that the last two issues that can occur with ML (unclear or incomplete data) are the ones where the greatest opportunity to refocus the machine lies.
This is where human-guided automation comes in.
If your data is unclear or incomplete, it is unfair to blame the machine. Neither human nor machine can make great decisions based on faulty data (though I suspect humans may happen to get luckier at times).
Where the Magic Wheel of Automation Comes Into Play
First, it’s not magic at all.
That’s my semi-dry sense of humor coming into play in the same way I like making fun of “hacks” that aren’t actually hacks – they’re simply best practice.
There is nothing magic with the wheel. You, the human manager, are simply working with PPC automation in a neverending process that includes these four (five, if you count the bonus) steps:
- Discuss strategically.
- Develop a plan.
- Build the plan.
- Analyze the data.
- Repeat endlessly.
When you have a specific problem you want to solve with automation, you need to begin by discussing what it is and how you will solve that problem.
This involves communicating with your client (or others who can inform this decision) and researching the different aspects of what you are trying to accomplish.
What sort of automated bidding solution will work best in this specific situation based on the data you have, your client’s objectives, and the goals you are trying to accomplish?
You must then put your strategy down into actual steps for which you can accomplish what you have created, and then you must build that plan.
Once everything is in place with automation, this is where many people stop: at the end of step three.
They pat themselves on the back as the automation is now at work.
But what happens if your strategy is missing a key component?
What happens if the automation doesn’t actually “work?”
Rather than stop here, you must now analyze the data you receive from your plan (make sure to give yourself enough time, of course) and most importantly, you must then be willing to move into a further strategic discussion to rethink and rebuild your automation.
Many people think because they are using ROAS targeting in this DSA campaign, that their job is done.
What happens if you rethink how the DSA campaign is structured?
What happens if you determine that this set of pages should be set apart and targeted with a CPA bidding strategy instead?
If you are managing automation well, not only are you never walking away from your automated campaigns, but you are constantly rethinking them and testing new structures, audiences, bidding strategies, and more.
Automation doesn’t take away your job; it changes your job.
Let’s play this out practically. Sometimes it’s easier to actually see it in a real scenario.
You decide to test Smart Shopping in this account, but it isn’t working.
You have tested your conversion data, uploaded complete and accurate audience lists, and have enough conversions over time for those products. Now, what do you do?
Time to start into Step 1.
You connect with the client or internally run an audit to determine what it is about these products that might provide a better subset of products to target.
Or perhaps you connect with the client to re-discuss the target audience of the different groups of products they have to offer.
This is a tricky step because it’s so important, but it can be done poorly without the advertiser realizing they’re not actually asking the right questions.
Regardless, let’s say you strategize with the client and determine that you have too many products advertised based on their budget.
You had one Smart Shopping Campaign set up to target 5000 products with a $10,000 monthly budget and it wasn’t working in this case. You take those learnings and develop a plan.
You are going to take their 100 most important products and group them wisely into two separate Smart Shopping campaigns based on categories (since the target audiences for those two categories differ widely).
You then build those two campaigns and push them live, ready to crack a cold one.
Now that you’ve pushed things live, you make minor adjustments with budget and ROAS targets, but you give the campaigns a few weeks to see what happens.
You begin to analyze the data, and you learn a few more things.
You learn that some of the products perform exceptionally well in this one category, and you end up pushing that campaign harder.
You have learned now based on this new test that the second category you chose is not a great set of products for Shopping campaigns.
Perhaps they are too expensive, or not visually appealing enough to get clicks with the dynamic ads in Smart Shopping Campaigns.
Regardless, whatever you learn, you decide to test another set of categories while leaving your first campaign running. You then talk to the client and begin to strategize the next test… and you are back into the wheel.
See how that works?
Unlike Thanos taking a break on his farm, your job is never done. Even with automation, you can’t rest (and hey, I’m excited about that!).
Human-guided automation is the future, and this is one way we look at managing it well in a process-oriented manner.
I hope this was helpful for you, and may the auctions be ever in your favor!
More Resources:
Image Credit
Screenshot taken by author, March 2021.
