As machine learning grows in pay-per-click (PPC), so does feelings of anxiety about future job stability. According to the Recruiter Nation Report by Jobvite, 69 percent of job seekers admit to being at least somewhat worried about losing their career to job automation. Man vs. machine is not just a common science fiction trope; it is evident right here in the reality of current internet marketing strategies. Many professionals in the industry see more of their duties becoming automated and wonder: Will I have a job in five years, and if so, what will I need to focus on to find career success?
The PPC optimization process can be broken down into a few interrelated areas that sustain each other, such as strategy, keyword research, bidding, and reporting.
The PPC Optimization Process
So, let’s compare how machines and humans perform at each step to assess where machine learning is gaining, where humans are losing and where humans are essential.
A successful PPC account starts with strategy. First, you must examine the business model. Identify what you want to accomplish with paid search, and then lay out a strategy to meet your goals within the business model.
Machine learning relies heavily on pattern recognition and correlations. Machines are terrible at coming up with ideas without information. They require a lot of input in order to function. However, humans excel at using prior knowledge to determine the proper action in any situation. That is why, as the MIT Technology Review reveals, humans still learn faster than machines. The realm of strategy lies firmly with humans and machines struggle to determine strategy. Therefore, in this area, human beats automation.
Listen to Brad Geddes' webinar on how to secure your PPC job in the face of increasing automation
Keyword research looks at your website, offers, the way people search, and comes up with a list of keywords to suggest how and where you want to advertise.
The Google Keyword Planner tool can scan your website and suggest ad groups full of keywords with just a few inputs. If the machine knows your website, it can do a lot of the keyword research for you.
Unfortunately, the machine doesn’t understand nuance. If you are a jewelry store, the machine might not recognize that you only sell high-end jewelry and not costume jewelry. The machine is good at making broad suggestions, but keyword research still requires human review.
Machines are making great strides at improving in this area. Dynamic Search Ads allow you to show an ad automatically if the search query matches your website. While the search ads might show for inappropriate queries at times, the machine has the capability to correct this as it gains more data. Once the machine has enough data to understand what searches will and will not convert a user to your website, the machine can fully automate your keyword research. The biggest problem is that this is a continuous process and you have to keep paying the PPC engines as the machine learns what keywords are useful for your business. Most people are not patient enough—nor do they have a high enough budget—-to waste money on a machine trying every combination while it learns.
This means that this area is not a total win for machines. Keyword research, therefore, falls into the realm of machine suggestion along with the human review.
Machines can write quarterly reports for public companies. They can take box scores and produce news articles. In fact, machines wrote over a billion articles last year.
However, a machine cannot write your ad copy. A machine can write based upon structured data, but it lacks the creativity to come up with original ideas.
Marketing is about connecting with people. One needs to be empathetic, business savvy and should have a good command over language to be a great ad writer. We’ve seen machine-written ads, and they are terrible. Humans beat machines in this area by a landslide. A machine lacks the imagination and emotional understanding needed to reach people. A great human writer is still highly in demand across paid search and will continue to be in demand until a machine can learn to understand what it means to be human.
Ad testing is important because it ensures you will target your market effectively and can save your valuable time and resources. There are four high-level components to ad testing:
- Creating variations of your current ads
- Coming up with brand new ideas for ads
- Determining that a test is taking place
- Determining when there are winning and losing ads based on data
Machines can only do two of these. Machines are very good at examining your current ads and suggesting tweaks for testing. If you use a different verb or adjective, switch up the lines within your ad and make changes based on your existing copy, machines can help by giving you suggestions.
However, when it comes to creativity and finding new ideas, machines lack the capability. As mentioned before, this is where humans outshine.
A machine can determine that you have two or more ads in an ad group and automatically start examining your data to determine winners and losers. A human should never have to waste their time doing something as simple as telling a machine that a test is running. So, using machines helps in streamlining the process.
A good ad test is rooted solidly in the collected data. A machine can automatically calculate minimum data, statistical significance and alert you when there are winners and losers. There are companies that do this automatically for all your ad tests today, saving you time and money.
Ad testing is an area that is both human and machine driven. Consequently, this area must be declared as a tie. Both are necessary for ad testing to succeed. The creativity and strategy come from humans; the data comes from the machines.
If there is one area that machines dominate, it’s bidding. Bidding involves taking a human-defined strategy and then changing bids based upon the target strategy, such as a target return on ad spend (ROAS) or cost per acquisition (CPA).
Ten years ago, a human was wasting a lot of time doing manual bidding or paying for a very expensive third-party software to do the bidding for them. Now, machines have taken this over. Most companies do not spend all day setting bids and can focus their efforts on other areas.
As bidding is based upon pattern recognition and statistics, this is one of the best uses of machine learning. Machines can easily determine how a user might interact with an ad based on previous behavior and the stats for that keyword, ad and landing page combination.
Reporting involves collecting data in order to gain a better understanding of your marketing strategies. There are three phases of reporting:
- Defining the report
- Putting the data into the report
- Interpreting the data
A machine is incapable of knowing what you want to know. Therefore, defining the report is up to you (the human). However, once the report is defined, the next step is easily automated because it involves plugging in the data on a regular basis over and over again.
Products like Google Data Studio are free and can automate your reporting once it is defined. If you need third party integrations for reporting, there are other companies that can fully automate this for you. Reporting is a repeatable task. Once the report is defined, a machine can do the rest.
Where a machine fails is in understanding how to interpret the data. A machine cannot take the set of data and build a story around it. It doesn’t understand the context of the data. While you should automate your reporting, this just saves your analytics team time in formatting the data. However, their primary goal should be to give you meaningful insights into what that data is telling you about your marketing efforts.
Humans are necessary for two out of the three phases of reporting, so they have the advantage in this area. However, machines are useful in automating tasks to enable humans to better and more accurately interpret data.
I’m writing this article in the month of March. A search for “Black Friday Deals” brings up ads touting sales that will occur on Black Friday, which is in the month of November. A look at the landing pages results in showing either an error stating the page no longer exists or a page that doesn’t contain any details about the deal.
A human knows that ads should not be promoting deals that expired more than three months ago. However, a machine may not. A machine needs to learn that a keyword or an ad is no longer performing, instead of being able to infer it is no longer relevant based on something as simple as the time period. As the searches for Black Friday deals have dropped, there are fewer data coming in for the machine to analyze. This results in the machine taking more time to finally understand that this term will no longer convert, and then adjust accordingly.
Eventually, machines will be able to use data collected over several years to figure something like this out. At the present moment, they have not reached such level of intelligence. When it comes to dramatic changes, trying something new or a seasonal period or sale has a hard stop date, it is necessary for humans to step in and instruct the machine on what to do. Man wins this round.
Overriding The Machine
If one of your landing pages breaks and goes down for a few days, the machine will see that your conversions have dropped and then lower your bids. If the bids get lowered enough, the ad will appear on page two. On page two, the ad will rarely get clicked, which means there will be little data coming in.
After the ad is placed on page two, you can quickly determine the page is broken and fix it. Unfortunately, the machine is not aware that you fixed anything, as it is still waiting for data to come in.
Since no data is flowing, the ad now sits on page two never to regain its previous position without human intervention.
There are times when humans must step in: When the machine’s algorithm breaks or needs to be reset, when data needs to be cleaned or if it starts finding poor patterns. The act of auditing and overriding the machine lies clearly with humans, as machines are not yet intelligent enough to self-regulate. Clearly, man has the advantage in this area, as well. Machines are not autonomous.
The Future of PPC and You
Through our examination of the various aspects of PPC, a few clear trends emerged:
Humans are great at:
- Data Interpretation
- Using prior knowledge
- Auditing the machine
Computers are great at:
- Recognizing patterns
- Giving answers within a predefined set of criteria
- Data calculations
- Repeatable tasks
If your job is centered around what a machine is great at, then your job stability is low and you should reexamine your future in PPC. If your job is centered around what humans are great at, then your job stability is higher and your future in PPC is promising.
Most jobs have an evolutionary path to them. While your current job might be what a machine is great at, you need to ask yourself how to transform it into a job where the human beats the machine.
For instance, if your job is putting together monthly reports—a machine can already do your job. Therefore, it is important to try to play up the strengths based on your humanity, and that would involve data interpretation. A good strategy would be to then evolve from someone who is creating reports to one who is interpreting the report data. Here, you can tell stories with the data so it can be used to form new creative strategies for the company.
Machines will only continue to get smarter and more powerful. However, they lack human characteristics. Marketing is about humans connecting with humans. Humans can be irrational, full of emotion and wonderfully creative. Embrace these characteristics.
If you make sure that your job is about connecting with humans, and showcasing what humans are great at, then you can beat the machine and enjoy a long, successful career in PPC.