21 PPC lessons from the age of machine-learning automation

You are not reading what I wrote. It’s a collection of PPC-specific lessons from those who do the work every day in this age machine learning automation.

Before diving in, a few notes:

• These are “lessons already learnt.”
• Things change. Platforms giveth, taketh and sometimes just change the way campaigns are managed.
• Below are lessons that combine strategic, tactical and “mindset-approach” based lessons.

Lesson 1: Volume is crucial to an automated strategy

It is simple: a machine can’t optimize towards a goal if it doesn’t have enough data to find patterns.

Google Ads might recommend “Maximize Conversions”, but the budget is low (sub $2,000/mo), and clicks are costly.

In such a case, you need to give it a Smart Bid strategy goal that can collect data to optimize towards.

You might also consider “Maximize Clicks” and “Search Impression Share”. This can be more useful for small volume accounts.

Lesson 2: Proper learning expectations

The second word, “learning”, is the key component of machine learning.

A machine must learn what works in order to learn how to do it.

This part can be very painful.

Expect to be underwhelmed by the initial Responsive Search Ad. To learn what works and what doesn’t, the system will need data.

These expectations should be set for yourself as well as for your stakeholders. These are the results of a real client example:

• RSA Month 1: 90 Conversions at $8 per.
• RSA Month 2: 116 Conversions at $5.06 per.

Month two was, as you can see from the graph, much better. Set the right expectations!

Lesson 3: Old dogs must learn new tricks

Many people who have been in the industry for a while didn’t learn how to manage ad campaign the way they should be run. It was quite a different mindset.

For example, I was taught how to:

• Instead of focusing on individual snippets, think about the whole ad. Then consider how they could be combined to create something unique.
• A/B testing with the Champion/Challenger methodology. The “control” is always a top performing creative. You can only modify one element at a given time, otherwise you don’t know what caused the performance shift. Machine learning can determine which creative snippets perform best for a subset of audience members and which one performs better for another.

Lesson 4: Keep up to date with site changes

Automation relies on the right inputs. Sometimes, a seemingly simple change can cause significant damage to a campaign.

Some of these changes include:

• Modify the URL of a “thank-you page”
• Additional call to action on landing page
• Plugins or code that slow down page loading
• Modification or addition of a step on the conversion path
• Replacing the hosted clip with a YouTube embed or Vimeo embed

These are just a few examples.

Remember, you can’t bet against site changes that occur without your knowledge.

Lesson 5: Recommendations tab

These are the best recommendations for Recommendations:

• Use your critical eye to evaluate them. This is a machine that doesn’t have the context you have. Take a look at the recommendations.
• Be careful what you click! It’s easy for a recommendation to be implemented, which is great, unless you accidentally click.

Lesson 6: Pay attention to Search Impression Share, regardless if you have a goal

Search Impression Share can be defined as “the number of impressions that you have received on the Search Network divided with the estimated number of impressions that you were eligible to receive.” It is basically a gauge that tells you what percentage of the demand for which you are competing.

This doesn’t mean that “Search Impression Share”, is the most important metric. You might also implement smart bidding rules with “Performance Max”, or “Maximize Conversions” which could negatively impact other metrics, such as “Search Impression Share”.

This is a good thing. However, you should be aware and OK.

Lesson 7: Keep up with changes (to the advertising platforms)

Sometimes things can change. It is your responsibility to keep up with it. Smart bidding is possible with Target CPA. It has been merged with “Maximize Conversions”.

Between July 2022 and September 2022, Smart Shopping and Local Campaigns will be automatically updated to “Performance Max”. These campaigns can be manually updated by you if you are running them.

Why would you do this?

• Automatic switchover eliminates surprises. Although there wouldn’t be any, it’s unlikely that there would be. It’s risky and not worth the risk.
• Reporting will be easier because YOU will choose when it happens so that you can properly note it.
• You can feel a sense of calm when you make the update happen at the time that suits you best.

Lesson 8: Keep separate records about your rules

This doesn’t have to be difficult. Just use your favorite tool like Evernote, OneNote, Google Docs/Sheets, etc. For each campaign, include the following:

• The what (goals and smart bidding rules, etc.).
• Why? (Your justification for this particular setup).

This is a great idea for three reasons:

• You are entitled to take a vacation at some point. The records are useful for anyone who might be looking at your accounts.
• You will be questioned about your approach at some point. Questions like “Why, exactly, did you set it up this way?” will be asked. It is a good idea to have the record handy.
• It will be easier for you to recall. It’s a win whenever you can get something out of your head and properly documented somewhere.

Lesson 9: Reporting doesn’t always translate into action

Imagine that you are setting up a campaign, and then loading snippets from an ad. You’ve got:

• Three versions of headline 1.
• 4 versions of headline 2
• 2 versions of the headline 3
• Three versions of the original description
• Two versions of the second description
• The list goes on…

Do you think it would be useful to know which combinations perform best in the given conditions? It would be useful to see if there is a consistent theme or trend. This knowledge could help you to create more effective snippets for your ads to test in the future.

It’s too bad, because that’s not what you get right now.

Lesson 10: Bulk uploading tools are your friend

If you have a large volume account that has many campaigns, you should be able to provide your inputs in a spreadsheet so you can do bulk uploads. You should also ensure that you have a quality control on any bulk actions.

Lesson 11: Automate mundane tasks whenever possible

A steady stream of mundane tasks can bring down morale like few things. Automate everything you can. This could include:

• Low performing keywords should be paused
• Ads that are not performing well can be paused
• Scheduling
• Bid adjustments based upon success metrics (example: Maximize Conversions).
• Bid adjustments to achieve the average position
• Bid adjustments during peak hours
• Get impression shares
• Controlling budgets

Lesson 12: Go beyond the default tools to invent

For an outsider, managing a large-scale PPC campaign for an enterprise would seem like having a lot of money to spend on high-volume campaigns. It’s a beautiful vision, but reality is often very different.

It can feel like managing 30 SMB accounts for those who manage these campaigns. There are several business units within each region, each with its own P&L.

You cannot exceed the budget. Period.

Also, ensure campaigns run throughout the month to avoid running out of budget on the 15th.

Below is an example of a Google Data Studio custom budget tracking report that shows the PPC manager how his budget is tracking for the current month.

Lesson 13: The 10% rule of experimentation

Try something new with 10% of your management effort (not necessarily budget).

If you have access to beta, you can try a beta.

Lesson 14: When you have to pin, “Pin”

You can place a “pin”, which will only insert your chosen text in the first headline. The rest of the ad will continue to function as a normal RSA if you are required to always display that message (e.g. by legal, compliance, or branding executives).

If you “pin” everything, the ad will no longer be responsive. It has its place, so if you have to pin something, you should pin it!

Lesson 15: The “garbage out, garbage in” (GIGO rule) applies

It’s easy: The ad platform will do the heavy lifting to determine the best ad snippet combinations that you submit to achieve your objectives.

The platform can do the heavy lifting to find the best combination between well-crafted ad pieces and garbage ones.

An RSA does not negate the need to have skilled copywriters.

Lesson 16: Educate compliance, legal, and branding teams in highly-regulated industries

If you have managed campaigns for an organization in a highly-regulated industry (healthcare finance, insurance, education), then you are familiar with the challenges that can arise. You are familiar with the frustrations and legal/compliance reviews that can occur.

Remember that you have your goals (produce campaigns which perform) and they also have their objectives (keep the organization in check).

Do yourself a favor and educate your legal, compliance, branding, and marketing teams about RSA campaigns:

• The high-level mechanics
• Benefits
• Drawbacks
• Control mechanisms available
• How it affects their approval process

Lesson 17: Don’t confuse automate with set and forget

To use an automotive analogy: Automation capabilities should be more like “park assistance” than “full-self driving”.

You might set up a campaign to “Bid to Position 2” then let it run without thinking about it. You may find yourself in a position 2 position, and a new competitor will soon enter the market. Now, you are running out of budget.

Automate the tedious tasks and do the heavy lifting (Lesson #11) but don’t forget a campaign once it is set up.

Lesson 18: Your business is more important than the algorithm

This is related lesson #5 and cannot overstated.

You might see a recommendation to reach additional customers for a similar cost per convert in a remarketing campaign. You can quickly spot inflated metrics in remarketing if you pay attention to the audiences being recommended.

You know the business better than any algorithm could. This knowledge can be used to guide the machine and ensure that it continues in the right direction.

Lesson 19: Some accounts may not be worth the squeeze

“Some accounts” is mainly a reference to low-budget campaigns.

Machine learning requires data, and smaller accounts don’t have enough activity for it.

You can keep those accounts as simple as possible.

Lesson 20: Look at what your peers do

Talk to industry peers and you’ll quickly discover someone who understands your daily challenges, and may have some solutions.

Network with others attending the PPC track at conferences. Register for PPC webinars, where tactical campaign management will be discussed.

Participate in (or just look) in social media discussions or groups that are specific to PPC management.

Lesson 21: Strategic PPC marketers are valuable

Many mundane tasks (Lesson #11) are now automated, so there is no need to spend hours doing them. This is a good thing, because no one enjoys doing most of these things.

Marketers who are only proficient at mundane tasks will be less in demand as automation continues to increase.

This presents a prime opportunity to strategic marketers to become more valued. It’s like the machine doing the heavy lifting requires guidance, direction, and corrective action when needed.

Marketers must be able to:

• A thorough understanding of your business objectives and key campaign metrics is essential.
• Assist the stakeholders of the organization in defining their overall business strategy regarding PPC.
• You can see how tactical capabilities can be used to achieve a specific business goal.

Search Engine Land’s 21 PPC lessons in the age of machine-learning automation originally published this article.