Should you be using automation & machine learning (AI) to manage your Google Ads campaigns?
Before we get to the answer, I think it’s important we take a trip down memory lane to help clearly distinguish between machine learning (artificial intelligence) and automation.
Automation
In recent years Google and Facebook have both invested heavily in automating elements of the advertising process. Initially these were basic features aimed at getting new advertisers onto their platforms more easily. Largely focussed on building your first ad campaign coupled with some basic objective setting.
The plan worked and Google rapidly expanded the number of advertisers on its platform. This actually presented them with a problem. They had a lot more advertisers, but these were marketers with much less experience and time on their hands to learn.
Google is essentially an auction business. The auction is for clicks (traffic or visitors) in exchange for money. The price is set by millions of bidders in the form of advertisers. Just like any auction, the more bidders you have for something the higher the price.
So google used automation to attract more advertisers, but this new batch of customers didn’t understand how to bid, or in most cases even know they were in an auction!
Not wanting to have a packed auction with no one raising their hands. Google decided to instead offer a helping hand in the form of some additional automation.
This time the goal was to help owners spend as much of their budgets as quickly as possible. Sorry, I mean help advertisers save time and effort by allowing google to bid on their behalf. And so was born automated bidding.
So you might be fooled into thinking this would be a disaster. But seriously what could possibly go wrong with combining inexperienced marketers with robots and credit cards.
Machine Learning
Machine learning unlike basic account automation is focused on data. In simple terms, the program learns from experience. The more data it receives the more it learns and the better its advice. At least in theory.
Google isn’t new to machine learning, they’ve been rolling out machine learning across business units since way back in 2011. The project is known as “Google Brain” and is absolutely nothing to worry about….. (It was slightly more concerning when they owned Boston dynamics robots as well, but thankfully they sold that business back in 2017!)
Machine learning has introduced some interesting features. Specifically Smart bidding (not to be confused with basic auto bidding) allows you to link conversion tracking to bidding. In theory, this allows advertisers to place a target cost per conversion on each ad campaign and then let google bid on each click as high as it likes until the cost per conversion level is reached. As its name suggests this is a much smarter way of bidding for each visitor, by forcing the advertiser to attach a specific goal to each campaign.
The second area that machine learning has impacted is ad creation itself. Responsive search ads have been pushed heavily by Google in recent months. And from next year responsive search ads (RSA) will be the only format available to search advertisers.
RSA’s allow Google to use machine learning to match your best performing headline with the best performing copy and calls to action with the right audience. It’s basically split testing on steroids.
Machine learning in particular presents some interesting opportunities for advertisers and is the area that’s getting the most attention from marketers. It promises the holy grail of online advertising. Set it and forget it marketing. If only it was that simple. Anyone familiar with more traditional online auctions like eBay and automation will already see where this is heading.
Let’s take a look a little closer at the pros and cons of using automation and machine learning on your google ads campaigns.
Pros
- No agency fees – You can potentially ditch your agency and let google do the account maintenance for you!
- Data-driven – machine learning is about as data-driven as it gets. Emotion (and some might say creativity) are taken out of the equation entirely. All bidding decisions are made my Googles brain, not yours.
- Improving ads spend efficiency – machine learning offers the potential to reduce waste and focus budget on the bids which are meeting your conversion goals most profitably.
Cons
- Lacks marketing creativity – A machine cant (at least yet) create an ad campaign concept. It can only deliver an offer to the right audience. You still need to tell the machine what you want to sell and why it would benefit the audience etc etc.
- Steep learning curve – Someone still needs to learn the system and load the “machine” with the right data.
- Still requires management – The machine still needs monitoring and adjusting based on results. Goals of campaigns rarely remain static. Seasons and stock levels for example often impact promotions.
- Trust – You’re placing trust in artificial intelligence to manage your marketing budget. (This is still very new tech)
- Requires data – Only works if you have existing data. Clearly the machine only knows what works when its seen what success looks like. This means you will need to invest in campaigns that don’t work (and manage them) to find the ones that do, before handing over the to the machine!
- Strategy – You still have to determine what bidding strategy to use, campaign structure, ad copy, landing page design, offer pricing etc etc.
- Commodity Risk – You run the real risk of turning your brand and its marketing into a commodity. Automating ad creatives, copy, bids all ultimately runs the risk of a race to the bottom when everyone uses the same method of bidding. A machine won’t help differentiate your brand!
- Brand message – You cant advise the machine to use your brand voice or messaging. It will simply use the combination of words that converts at the highest rate. Regardless of the story this tells your new customers. Commodity VS Brand is the issue yet again.
Recommendation
If you’re selling a commodity that requires little differentiation and has no story attached to a brand, then automation could be the right call. There’s still a considerable amount of time and management required, but the potential savings at least in the short term could be substantial.
Overall the automation trend offers some potential short term advantages to early adopters. Much in the same way that auction sniping software was effective on eBay until everyone began using the same tools, the advantage will be short-lived. More worrying however longer term is the realisation that Google ultimate aim is to bid against itself using your credit card.
Google is not alone in this strategy. All the major platforms have extensive machine learning rollouts. Facebook in particular offers an even more comprehensive array of automation options with the same goal in mind. Turning brands and advertisers into commodities sold off the back of a series of ones and zeros.
Brands will increasingly need to differentiate themselves to set themselves apart from their competition or run the risk of becoming commodities whether they like it or not.
