A two-part seminar on automation in paid search; Optmyzer’s Frederick Vallaeys and Adalysis’ Brad Geddes outline the importance of automation in PPC, and how human work can work in tandem and support this rapidly growing aspect of PPC.
Frederick kicked off the session by asking the ever-important question: Why do we need to automate?
- To keep up with your competitors: Automation frees up time to do more strategic work. It allows companies and individuals to scale up and manage more accounts per person, leaving more space for growth.
- To change the rules of AdWords: Sometimes, Google drops a big game changer eg. Google now allows campaigns to spend up to double of their daily budget based on past trends; this might be great for the big spenders, but for SME’s with a limited budget, this can be detrimental. Automation can put you back in charge – A simple script can automatically stop any campaigns after they’ve spent what you intended them to, and re-enable them the next day.
For these reasons, we need to embrace automation and learn to work alongside it, as AI innovations are accelerating – We can’t expect to be given a living for doing something a computer could do, better.
This leads to a second question: What roles will we play?
- Prevent automation from going wrong. There are scripts and tools for monitoring accounts, for example, the anomaly detection script, which allows us to change sensitivities and aspects of the campaign to check. Another example is the broken link checker, which detects 404 errors on landing pages. Our role is to monitor these, making changes ourselves when the account fails to run itself, with help from automated alerts.
- Improve results by adding human judgement: An example of this in action is the query management script. This reviews search terms and automatically adds them as targeted keywords if it’s a top performer. However, a search term with fantastic metrics could be added as a keyword, but this includes your competitors brand. The term performs well for the competitor but will perform very differently for your company. Automation saves time finding opportunities, human judgement prevents blow-ups.
- Train your own Machine Learning model: Improve results by leveraging better data, perhaps even your own data. For example, mixing automated CPA bidding strategies with your own margin data provides Google with more metrics to work with, giving the AI a better chance to succeed.
- Solve problems a different way: For example, a script that chooses which ad is better might just turn the ad off, based purely on data. But what if the reason it isn’t performing well is that there was an error in it, like an incorrect year or offer in the headline? The reason for underperformance may be a lot more obvious to human eyes.
- Decide what to automate: Things you do often and take a lot of time. Eg. bid management, budget updates, ad testing, how long and how frequent does it take/do we do them? Prioritising areas of automation can be very important to your PPC strategy.
- Automate the tasks where steps are well defined: can you write it down as a system? For example, low impression share & low CPA = automated rule to raise bids
- Break down complex tasks into automatable parts. Eg. ad testing and writing new ad copy: 1) automate to find the losers of split tests and turn the ads off 2) automate to find the winners and use them as the base 3) automate to find the best CTA 4) The human touch: combine CTA’s with new ads
Brad Geddes then took to the stage to further explain the role of humans in a machine-dominated world. He outlined three levels of automation: Manual, Computer Assistance and Full Automation.
The advantages of doing everything manually:
- Amazing amounts of control
- Human Intuition
Whereas, the disadvantages are simply that this entails more, often boring, work.
Outlined below are a few examples of manual work having an advantage over automation:
- If an ad was displaying an out of date offer, a computer might not stop running it, but as humans, we have prior knowledge of what will and won’t work, and if there’s no data for the computer to work with, it can’t do anything. Obviously, an out of date ad won’t drive large numbers of sales.
- Manual work can avoid what is known as “Automation danger.” For example, delays in conversion tracking, broken GA code and website downtimes could add a vital keyword like a Brand as a negative keyword through automation.
- “Alert danger” can also be avoided. Automation might give a lot of alerts that don’t mean anything particularly important, so we may delete them without looking at them. Here, automation is giving ourselves more work
So when should we think about automation? The answer was simple: Can we flowchart it and be happy with every single decision? If so, do it. For example, its possible to see what the outcomes of automated bid adjustments will be, so these should be automated.
Brad moved on to asking how automation will affect our jobs: Automation creates or shifts job responsibilities – jobs are lost but more new jobs are created.
For example, the rise of the bank teller: When ATMs became popular, bank branches needed fewer employees due to the automation of withdrawals and deposits, but as it became cheaper to operate banks as a result and everyday tasks become less time-consuming for consumers, they had more intelligent questions to ask, so more banks opened as a result.
We can apply this to Paid Advertising, too. Market research, sales orders and stat assistants will be made redundant if their jobs don’t evolve into someone who interprets research, stats or provides sales strategy.
The session ended with the question: Who wins? If the tasks included only creativity, strategy, storytelling, empathy, then humans. If it includes only repeatable data, stats, and pattern recognition, then computers win.
A combination of the two, or “Computer Assistance,” is vital to ensure that marketing campaigns can be run effectively, whilst being given the chance to grow whilst keeping an eye out for those mistakes which can be detrimental.