It’s no secret that AI and machine learning is topical right now in the PPC world with several large agencies bringing out their own AI and machine learning tools, with some even promising that they outperform humans.
In fact it is becoming commonplace for agencies to claim that they are driven by AI, but in reality this is very often not the case.
So is all of this new technology worth the hype and can AI really outperform humans when it comes to running PPC campaigns?
In this article I’m going to try and do two things, firstly take a look at technology that is behind these futuristing AI driven tools and break down how it works.
Secondly I’m going to argue that machines are better than humans at PPC (about 80% of the time) and I’ll also discuss what’s best to do with the other 20%.
(N.B. Econsultancy subscribers can download A Marketer’s Guide to AI and Machine Learning)
The current AI landscape
An expert from the field of AI could argue that paid search agencies don’t use AI at all. In the vast majority of the cases where paid search agencies are using AI it has been around bidding algorithms.
Which in reality is simply “machine learning” a subset of AI where a computer uses a dataset to iterate and progressively improve on a specific task as opposed to “deep learning” where computers are programed to learn in the same way humans do.
In the case of bidding it is usually used to either change bids based on position where a machine learning algorithm will iterate towards the most profitable position. Or alternatively portfolio based bidding where sets of relevant keywords will be grouped based on common attributes and an algorithm will iterate towards a defined goal, prioritising the lowest cost conversions first.
Machine learning has also been adopted to analyse data and create models. Here is a great example of a machine learning script by Russell Savage that allows advertisers to see how the weather affects their campaigns.
There have even been experiments with totally replacing humans with robots in the case of Albert. But in reality they are still no match for the flexibility, strategic ability and intelligence you get from a human account manager.
Overall the current landscape has not changed much over the last decade looking back when the early machine learning based bid platforms were released and its years, probably decades away from the first fully automated AI driven agency where robots will replace humans.
So where should advertisers and agencies be focusing their efforts to make their media buying more effective? The answer is automation.
Automation and the shifting role of the PPC manager
In the near future it is much more likely automation will start to replace a lot of the manual work that media buyers do on a day to day basis that doesn’t require any higher cognition.
When you look at what a PPC manager does on a day to day basis it quickly becomes apparent that around 80% of what they do can be automated.
Take a task for example like setting bid modifiers at device level for a set of 100 campaigns.
This task would usually require a PPC manager to export the device level report, use excel to calculate the bid modifiers and then manually go and enter them into Adwords editor which would take about an hour.
However using the Adwords API or scripts tasks like this can be done instantly across a large portfolio of campaigns.
It reduces the amount of human error that is likely to occur as well as allowing for this task to be done hourly even when the PPC manager is sleeping.
This is a great example of tasks that is much better suited to machines than humans.
But the lists go on. The table below breaks down the normal set of tasks you would expect a PPC manager to do daily or weekly and identifies which ones are better done by machines than humans.
As we can see from the table above automation really comes into its own where there are either large datasets or highly repetitive tasks.
Contrast that with tasks such as writing ad copy that people want to click on or understanding a client’s needs and creating a winning strategy around those needs. This is clearly where humans come into their own.
So in conclusion, automation and machine learning technology will likely do more and more of the heavy lifting that PPC managers do (reluctantly) day to day. Freeing up their time to work on more important tasks around strategy and working alongside technology to continually split test ads that will resonate better with their audiences.
In conclusion although a large number of agencies are claiming to be AI driven the technology behind them is in almost every case “machine learning” as opposed to “deep learning” which is what most people would associate to be AI.
The agency of the future where robots run campaigns and can write ad copies that outperform humans, is at the current rate of progress, decades away.
Until then automation is a great way to replace a lot of the mundane heavy lifting work that paid media buyers do on a day to day basis, freeing up their time to focus on more important higher value tasks like strategy and competitor analysis.