PPC pros who underestimate the impact automation will have on their professional lives risk missing out on a great transformation.
I’ve been writing about the ways automation is transforming PPC for a few years now.
Through that time, it’s become abundantly clear that managing PPC is rapidly moving away from requiring PPC tacticians.
We used to operate in the details, such as focusing on exact keywords or specific bids.
Now PPC pros find themselves managing such things more indirectly – as if we’re shifting more to the periphery of the ads account.
Google’s machine learning is the automated box that keeps swallowing up more of the tasks we used to do.
And every time it gobbles up another lever, we have to shift our own work to that of the PPC teacher by changing a setting or input at the edges of where our control ends and the world of machine learning starts.
There are three key factors at work in expediting this transformation of the PPC manager’s role.
Factor 1: More Automation
We all see the frequent updates from Google about new forms of PPC automation that are supposed to make our lives easier and deliver better results.
For example, they recently announced the limited beta of Performance Max campaigns, a new campaign type that will serve ads across the entire Google Ads ecosystem.
Previous automated campaign types were more limited in terms of where they could serve ads, and advertisers had to create several automated campaigns to reach the largest possible audience.
What’s driving the onslaught of more automation are the improvements in machine learning that are themselves driven by two key factors:
- More data.
- Faster processors to find signals within this data.
I’m always fascinated when looking beyond big announcements and I hear about the incremental changes enabled by better machine learning.
Small changes that, when aggregated, eventually turn even skeptics into fans.
For example, Google recently announced that Smart Shopping campaigns can now be started even without the addition of a conversion tracking tag or remarketing tag.
They specifically noted the following:
“Finally, to make it easier for new retailers to get started, we’re reducing the eligibility requirements for Smart Shopping campaigns. Though a remarketing and conversion tag are still needed for optimal performance, you’ll be able to create your first Smart Shopping campaign now and take care of these tags later.”
Considering that this campaign type works only with automated bidding, specifically “Maximize Conversion Value” with an optional target ROAS, it’s fascinating to see that conversion tracking is no longer needed.
What’s most likely going on here is that Google’s machine learning has gotten so good that it can draw assumptions about any advertiser’s performance.
This is based on signals from similar advertisers who do have all the measurement tracking pixels in place.
Advertisers have benefited from machine learning’s ability to predict similarities between users.
Now Google is turning the tables and predicting similarities between advertisers.
When they can predict any advertiser’s likely conversions, they no longer need advertisers to report this data to enable automatic bidding.
What We Can Do
Google itself says that for Smart Shopping campaigns, conversion tags are still needed for optimal performance.
Of course, the machine’s predictions will be better if it has data from that particular advertiser to base its decisions on.
What we as advertisers should realize is that Google’s goal is maximizing conversion value from the campaign.
And most advertisers probably care more about maximizing profits than revenue.
To achieve more profit, advertisers can either report profits instead of sales value through the conversion field or they can create multiple Smart Shopping campaigns, each with their own tROAS.
In fact, we’ve recently started hearing this last piece of advice from Google reps more frequently.
They’re basically saying that one way to optimize performance is to stop treating the entire product catalog as a portfolio optimization problem.
Instead, group products that are similar in profitability and then manage each of those subsets as a portfolio by grouping them into their own campaign.
Factor 2: Less Data
We’re constantly asked by Google to change the way we optimize our accounts when they take away a piece of data we’ve come to rely on.
While Google’s machine learning is benefiting from the ever-expanding trove of auction and user data, advertisers who hoped to do something smart with data on their own are seeing the firehose turning into a garden spigot.
Most PPC managers I know consider managing search terms a key way to optimize accounts.
Both to reduce wasted spend from weird close variants and to build volume by capitalizing on trending new searches before everyone else does.
What We Can Do
While we may not like it, the reality is that every time we lose some data, it’s the experts who figure out and implement workarounds to continue having an edge over less savvy competitors.
For example, while we can still look at search terms the same way we did before, we may just get fewer ideas for optimizations now that the data has been curtailed.
But we can also make some assumptions about the statistical distribution of the data and use different techniques like n-gram analysis to find optimization ideas.
This technique was proposed at my recent PPC Town Hall with Martin Rottgerding and Brady Cramm.
The idea is that the same combinations of one, two, or three words (unigrams, bigrams, and trigrams) that we no longer see as individually reported search terms might still surface during an n-gram analysis and can hence be the basis for a decision about a new negative keyword to add.
This is a less direct way of managing the account, but another example of how Google is forcing us to manage PPC accounts in a more indirect manner.
Factor 3: Fewer Controls
This factor really goes hand-in-hand with the other two.
Having fewer controls is almost the inevitable result of less data and more automation.
For example, Smart Shopping campaigns automate bids across dynamic remarketing and shopping ads, and across search, display, Gmail, and YouTube.
However, they don’t report this level of detail about where the ad was shown and how it performed there.
As a result, having the control to manage this would make little sense.
While we could make gut-based decisions, we couldn’t truly optimize what we haven’t been able to measure.
The newest automated campaign type, Performance Max, is similar in how it works.
It can show ads in even more places, such as Discovery campaigns and in other formats, like video.
Bidding is automated and likely won’t be something that we can control.
The only thing we may have control over is the message.
Or at the very least, the components that should constitute the message, like what we have with Responsive Search Ads (RSAs).
What We Can Do
We long may have been overly focused on the things we can do to optimize PPC in spreadsheets: calculate bids, find targeting ideas based on numerical analysis, etc.
Connecting with our prospects through a really compelling message is often left to the team that does CRO.
Crafting a compelling message that connects with our prospects and tells them why ours is the business worth considering to solve their needs is something that we will need to focus on more.
Luckily this is still something we can do with our spreadsheets, at least to some degree.
The difference is that unlike with managing bids and search terms, our spreadsheets don’t tell us the final answer.
They just point us in the direction that needs work.
And then we have to do that work of being good marketers.
I continue to believe that machines alone won’t deliver the best PPC campaigns.
The machines can help Google grow its advertiser base by delivering quite adequate results to PPC novices.
But the best campaigns can only happen when PPC experts tap the power of the machines while adding a human element of strategy and insight.
How we harness the power of the machines is shifting due to the three factors I covered here.
As a result, we’ll need to be ready for a world where we manage fewer of the details and act more as the teachers to the machine.