PPC automation will disrupt your business, but automation layering will save the day

Considering how much Google Ads has changed in the past decade, it stands to reason that successful account management should also have evolved dramatically.

My friend and industry pioneer Andrew Goodman recently put this idea of agencies not keeping up with the times in slightly different terms and wrote that self-proclaimed PPC experts don’t keep up with the nuances of all that’s changed in Google Ads because they settle for good-enough rather than going for greatness.

So what follows are my thoughts on how to whip your PPC management skills into great shape for 2020 and beyond. The core idea is that using automations from Google is inevitable, so if you want more control, you can regain it by layering your own strategies on top of Google’s through a concept I call automation layering.

It’s not a question if, but when automation will do it better

Back in 2007, Google launched Conversion Optimizer which helped advertisers with at least 300 conversions over a 30 day period reach a CPA target for their ads. Today that strategy is called Target CPA and requires 20 times fewer conversions (15) to achieve similar results.

There are now also about 11 types of bid management strategies to choose from. Some, like Target CPA and Maximize Conversions almost seem like different flavors of the same thing, showing that things have progressed far enough that the differences in strategy can be down to nuanced differences in goals. It takes serious know-how to make sense of all those strategies and how they interact with manual controls such as bid adjustments.

The point is that there has been a lot of progress in automation for bids. And that’s not even to speak of automations in creatives, targeting, etc.

The creep of automation into all areas of PPC will continue unabated due to two driving forces:

1. An ever-expanding trove of data. Every time a search happens, Google learns a bit more about the correlation between users and what ads resonate with them. The more data one has, the better the signal gets.
2. Moore’s Law and the seemingly endless cycle of computing power doubling every 18 months. As computers can do more calculations with the vast data from point 1, they can make better predictions. And while even Gordon Moore believes his law will hit its limit by 2025, there are already teams working on ensuring sustained growth in computing power.

So please, figure out how automation fits into your business plan. Too often I hear it’s something people are too busy to figure out right now. “We might get to it another day,” they say… Well duh! If you used some automation you might actually have time to get strategic about your own business, and maybe even for the clients who pay your bills.

The time to prepare is now

Even if you don’t live in California like I do, you’ve probably read that PG&E, the utility company, turned off power to 800,000 customers to prevent catastrophic wildfires due to aging and poorly maintained transmission lines to avoid the type of wildfire that destroyed the town of Paradise in 2018. What does this have to do with PPC? Well, it showed me how bad people are at long-term thinking and planning.

Now that the power’s gone out, all of a sudden everyone’s upset at PG&E because households are in the dark. And while the utility company is certainly not blameless, they’ve been advising customers for months that this likely would happen. But until the warnings turned into reality, few consumers took time to prepare.

Google executives, industry bloggers and industry peers have all been saying that automation is coming to disrupt us. And yet too few agencies heed the warnings and will wait until it’s too late to change.

Complement rather than compete with automation

So if it’s inevitable that automation is here to stay, a modern account manager better figure out how to make it part of their routine. But if for some reason, you still believe that you can compete against automation and win, listen to this advice from Hal Varian, Google’s Chief Economist.

He said:

“If you are looking for a career where your services will be in high demand, you should find something where you provide a scarce, complementary service to something that is getting ubiquitous and cheap. So what’s getting ubiquitous and cheap? Data. And what is complementary to data? Analysis.”

You know what’s getting ubiquitous in PPC? Automation. Many of these automations are not just cheap, they’re free and available to all takers in Google’s own Ads management interface. Sounds exactly like the type of thing Varian was talking about. Heeding his advice, to me it seems smarter to become a great complement to automation than to fight it by continuing to do the same old things manually.

Automated tools don’t guarantee success – you do

The reality of PPC automations is that results vary from one advertiser to the next so not everyone will see case-study-worthy results when deploying the latest Google smart feature. But on average, automated tools deliver better results with less effort, so if you find yourself as an outlier who’s not successful, you have to first determine why. Then you can plan a strategy to get on the right side of the averages.

For now a lot of the automation we see solves very narrow problems. There is one system for automating bids, another for optimizing ads, and yet another for matching those ads to likely prospects. The role of the agency is to piece the right solutions together and ensure they work well together. For a more in-depth look at this, check out my post about how the wrong combination of automations can destroy an account.

Through the example of making the mistake of using an attribution model like last-click that won’t work well with other automations, it should be clear that the results of automation can be extremely dependent on the knowledge and skill of the account manager. If you don’t see the results others are achieving, consider that the fault may not be with the tool but with the person using the tool.

As someone who builds PPC tools for a living, I know that even if the user is to blame for bad results, it’s still the tool creator’s problem. A tool that is amazing at automating ad testing is useless if it’s too difficult for the average advertiser to use without making mistakes. But the best advertisers don’t let shortcomings of technology get in their way and will work to produce amazing results with what’s out there.

Automation doesn’t mean set-it-and-forget-it

There’s this interesting cycle that tools from Google often seem to go through. They start out manual, then become automated, and eventually, some new controls are added. In essence, this turns the automation into a new type of manual tool.

Bid management is a good example. First, we managed CPCs and bid adjustments manually. Then it became automated as a smart bidding strategy like target ROAS. But then Google added the ability to set ad group level targets, mobile bid adjustments, and seasonality bid adjustments. If you will, the automated target ROAS bid strategy is pretty manual if you consider all the settings you can now control.

Just the mere fact that there are several settings and controls for automated tools should be a dead giveaway that the automations can be optimized if advertisers are motivated enough.

I recently asked advertisers the last time they had an account with a single CPC bid. Or the last time they had manual CPC bids that they never changed. The answer for most was: “never!” Target CPA and target ROAS goals should likewise not be the same for the whole account, nor should they be static. They should be managed for better results, and certainly not treated as a set-it-and-forever-it tool.

Humans can make sense of the data

Curious why I say target CPA and target ROAS should be monitored and managed rather than left alone? We should do this because there are factors in everyone’s business that affect conversion rates that Google’s prediction systems may not be picking up on. The automation only detects a change in metrics and in its narrow scope of what it can do, may very well use this data to do entirely the wrong thing for the business.

Here’s an example that I’ve personally encountered… Automated bidding one day noticed that conversion rates dropped significantly and so bids were reduced in an effort to maintain the target CPA. As a result of the much lower bids, the advertiser’s conversion volume went off a cliff and never automatically recovered.

Here’s what happened. The bidding automation correctly saw a decline in conversion rate and adjusted bids downwards. But as a dumb automation, it never asked why the conversion rate dropped. The humans knew it was because a new landing page launched. Humans would have known the correct response to this event was to revert back to the previous landing page rather than decrease bids.

Machine learning is bad, very bad at explaining why and how. Show it a picture of a cat, and it knows it’s a cat. But good luck getting it to explain why it’s a cat.

In my book on the future of digital marketing, I explain that one of the roles humans need to play is that of PPC pilot, someone who monitors the automations and can make course corrections if bad data is causing bad actions to be taken.

Understand what the machine is doing and make sense of the data it is using to make its decisions. That is pretty much what Hal Varian said in the quote I mentioned before.

If automation isn’t working or you don’t trust it, consider automation layering

If you believe that automation will become even more pervasive in PPC in the future, then it makes sense that we must learn how the automations given to us by the engines work so that we can optimize them by managing their settings.

It’s the simple premise that humans + machines are better than machines alone. But I think in PPC there’s an additional meaning to that premise. Perhaps the equation should be:

“Human (account manager at my company) + machine (automation created by Google) is better than machine alone.”

An advertiser’s worry is not just that the machines are taking over our jobs, but it’s that the machines are built by Google, who also collects many of our advertising dollars. As an ex-Googler, I trust that Google tries to do the right thing, but there’s nothing wrong with wanting some guarantees and putting oversight in place, especially when even Googlers can’t really explain exactly how their machine learning automations are arriving at their decisions.

So that’s why PPC managers want to be in the equation rather than letting machines do PPC on their own. Humans can monitor the machine’s decisions and provide corrections and guidance when those decisions appear sub-optimal.

Script your PPC process for better results from automation

But adding manual human labor to the mix is counterproductive to efficiency and economic growth. Automation would be better. So what account managers should really strive for is to have their own automations to monitor and optimize the automations from the engines. That’s automation layering.

So instead of striving for:

Human + machine

We should strive for:

Advertiser controlled automation + engine controlled automation

But as enthusiastic as marketers are about building their own automation (thanks to everyone who’s been downloading my scripts over the years), the truth is it doesn’t always come naturally. That’s when third-party tools can be helpful.

They help advertisers control what automations do to their accounts. I’ve explained how to use automation layering to monitor and control close variants that can muddle the meaning of exact match keywords. The following graph tries to explain it more conceptually but check out my previous post if you want a more tactical guide.

Close variants allow Google to target ads for exact match keywords (the small circle in the middle) to a much larger set of search terms (the big outer circle). Google is in control of how big they make this outer circle. If they want more revenue, they can literally change some settings in the code to make the circle bigger to make ad auctions more competitive. Advertisers want more certainty and control. So with automation layering where they control the automation, they can scale back the search terms (the dotted circle) to a level they feel comfortable with.

The bottom line

Automation from the engines is already disrupting PPC agencies and will continue to do so more. Anyone who’s been coasting and doing PPC like it’s 2008 needs to come to terms with the fact that this is not a future-proof strategy. Figure out how to bring automation into the mix of what you do. And if you haven’t because you don’t trust it, know that techniques like automation layering can restore some level of control.