Automatic? Yes. Supersonic? Sometimes. How AI can be used to drive PPC automation

22nd August 2018 - 6 minutes read  PPC
Luke Townsend - Marketing Director

In the beginning (or in the mid-1970s), there was the personal computer. Then came the internet, and today, we’re on to the next big technical innovation. It comes in the form of artificial intelligence (AI).

We know that most fundamental digital marketing activities rely on automation, which are ultimately driven by AI. And as technology develops apace, pay-per-click (PPC) marketing has become eligible for an automation transformation of its own. AI is revolutionising the way in which digital marketers allocate time and energy, freeing them from the banal and repetitive, and handing those tasks to our faster friends, the computers.

There are plenty of methods for achieving AI.  It’s true that you could – if you had the time, energy and temperament – manually program AI by writing lots of code made up of “if-then” statements, for instance. In the PPC world, that would be equivalent to using Automated Rules to fully codify how your account manager optimises an account (i.e. not very efficient). Machine learning is a better example of how we might make the best use of AI, and is the statistical approach that has proved so useful with AdWords – it helps predict which ads are most likely to attract clicks.

Auto-horror: the dark side of automation

When AI is managed badly, it can pose serious problems for PPC campaigns. These errors are entirely possible, so keep your eyes peeled for the kind of mismanagement that might lead you down automation holes.  

An automated search query, for example, might select new keywords without checking that they’re contractually viable – in other words, they’re prone to the kind of errors humans are less likely to make.

Similarly, if you’re using systems like Google’s Flexible Bid Strategies to achieve a satisfying return on advertising spend (ROAS) for AdWord campaigns, say, it’s possible that you won’t collect enough data at the right time in the right place. You might find that keywords are bid down to lower pages, but the change won’t be flagged in time to correct the misstep before conversion levels drop. We set up automated systems to increase efficiency, but we also need to know how closely to monitor the automation in order to keep them on the right track.

How can automation safely drive PPC?

Of course, if you’re time rich and don’t mind taking things slow, you’re welcome to go it alone – take the manual route, and settle down with our old friend The Spreadsheet.

If you don’t want your paid media campaigns to make your life an extended living hell, though, most would recommend you take a step up to the first stage of automation and get acquainted with tools like Google’s Anomaly Detector, which flags up deviations from expected performance on an hourly basis. The next stage would be automating a selection of individual management tasks, but crucially without connecting them. These kinds of automations can run for 24 hours and are there to pause any keywords whose Quality Scores drop below parameters set by the user.

A more comprehensively automated PPC tack would allow automation to manage a range of interlinked tasks and components. Bids and budgets might be arranged to encourage interplay, for example: if bids are raised, the system could reallocate budget to the best-performing campaigns first.

But all these levels of automation require regular human supervision. It is possible to build a campaign that has established goals and a vertical platform so clearly defined that keywords, budgets, ads, bids and targeting options can all be managed by the machine. Full automation – the stuff of Google’s sci-fi dreams – negates the need for human input altogether. But don’t jump the gun (either in excitement or fear) because full automation remains a distant reality for the time being.

We need to be vigilant about how we use AI to drive PPC automation because if it’s not done properly, we can do damage to our own campaigns and the way we develop systems as a community. It’s important to have open conversations about how users are putting the technology to work, and what we can learn from their experiences.

PPC is going to continue to change radically and rapidly. Automated PPC will become more prevalent, but it will be more sophisticated, too. And in any case, humans will always be there to monitor automation, because let’s face it, it doesn’t matter how finely tuned PPC campaign is if you can’t get anyone to convert.