If you are managing a Google AdWords account (or a bunch of them) and are tempted by the recommendations that Google AdWords offer, slow down. Don’t accept these recommendations blindly assuming they will work for you simply because Google is telling you they will.
Google is using more artificial intelligence (AI) all the time, investing heavily in machine learning. Their AdWords platform is certainly part of this growing revolution, actively learning and becoming better at detecting optimisation targets and trends.
The problem with automated suggestions is that in many cases, it isn’t that effective. Not yet at least.
Google AdWords’ automated recommendations may have a net positive result when viewed from Google’s lofty heights, but as individual account holders or client managers, we can’t afford to be one of the losers.
I have tested my own strategy against their automated ‘recommendations’. In using the AdWords recommendations I achieved two things:
- Poorer results.
- Higher costs.
These are the natural enemy of any marketing consultant or manager.
While I could outline every concern that I have based on my experience, I don’t want to come off as negative towards Google Adwords. I’m not. It is an extraordinarily effective marketing platform and should be part of any digital marketing strategy. That said, the AdWords automation tools need further refinement before I will place any trust in them.
AdWords recommendations I really don’t like
Of all the suggestions AdWords have made via the Recommendations tool, several have been stand-out failures in my testing. They are:
- Using Google search partners.
- Anything to do with automated bid strategies.
- Creating dynamic search ads.
Their other recommendations may or may not work for you, depending on your campaign set-up and goals, however, I’d be very cautious when using those listed above if I were you. If you do use them, monitor them closely and often.
The search partners websites simply don’t return the same results as the Google-owned properties. You don’t even know which websites are acting as search partners – who exactly are the “hundreds of non-Google websites“?
Then there are automated bid strategies. I am naturally tentative towards changing bids automatically. Sceptical even. But a friend of mine who also manages a select list of clients said he had some success with automated bid strategies. So, I gave them another shot. Unfortunately, the automated bidding strategies only served to drive up costs while producing fewer conversions. My hand-crafted campaigns and their tightly controlled bid strategies always beat the automated alternative.
And then there were dynamic search ads. While they can certainly save time on large and very large accounts, they tend to give ads a generic taste. I don’t think you should rely on them either unless you simply don’t have the time to give your campaigns the human touch. I don’t use them at all currently.
My approach remains unchanged
As a professional AdWords consultant, I need to be right on the ball. Failing strategies are not acceptable to my clients and so I, for one, will not implement a recommendation just because it has appeared in the recommendations section of the AdWords platform.
While I’ll always keep a close eye on the recommendations that AdWords make, I am in no rush to try them again any time soon.
I spend a lot of time manually controlling my client accounts to ensure they are optimised. This means I don’t rely on tools created by others to run algorithms against the accounts that I manage. Yes, I use some scripts that allow me to be efficient and more effective (I used to be a programmer so this is easy for me), but my client accounts will continue to be governed purely by human intelligence.
It is no secret that Google employs some of the smartest people on the planet and have the resources to solve just about any technical problem mankind is capable of solving. Therefore, I have no doubt that the artificial intelligence used on the AdWords platform will continue to make huge improvements. At this stage, however, it still seems to be in the early stages of its learning curve.