Automation is here to stay and, as Paid Media marketers, we must learn to embrace it. I explain how automation and machine learning has changed and what we can expect in the future.
Today, I’m going to be talking about automation and the changing face of media. Specifically, the trends that we’ve seen in updates on products released by Google using machine learning algorithms and AI, and the way that has shaped the tools and platforms that we use as marketers.
But, before we get stuck into that, just to do some scene setting, we’re going to go back to October 2000, when AdWords was launched. If you were lucky enough to be working in paid search back then, this is the interface you had to use. It was much simpler back then, much less data to optimise against. There was no way of doing bulk changes, you had no complex product feeds. Because of that, campaigns tended to be a lot smaller and a lot easier to manage.
Contrast that to today, and the amount of data that Google has access to at its fingertips, things like device, location, browsing habits, previous search terms, salaries, job titles, other demographics data. That’s not even really scratching the surface about what they know about you. We can use all of that information to target users and to adjust bids, to change targeting, to be more accurate, and to be smarter about what we’re doing. But it’s a big task to do that granularly and manage all of that data. Which is why Google have released automated products and machine learning products that can do a lot of that for us.
So just a couple of those, just a handful of things that they’ve released in the past. Things like smart display, which will automate the targeting and the creative that’s used in display campaigns. Bidding strategies and double clicks that will take control of all of the bids that you’re placing based on what it knows about the user. Even dynamic search ads that’ll automate some of the ad copy and decide what keywords to target based on your website copy. Budget plans, automate pacing and distribution of your budget.
Now, there’s pretty much a fully automated workflow for every channel you could want to advertise on. Google aren’t slowing up either. Myself and Deyna went to the Google product kick off meeting last week. And out of 45 new updates that they announced, 27 of them involved machine learning or automation. And that’s across all of their products. Over half of them involved machine learning automation.
So why are they doing this? There’s a couple of reasons. Firstly, as a human it’s impossible to make use of all the data we have access to today, processing every piece of information and making decisions based on what we see. It’s just not efficient to do it manually. And if programmed correctly algorithms can make the same decisions in a fraction of the time. And because of that, they can perform better. It’s not guaranteed that they will perform better – they’re not faultless – but they have potential.
And secondly, if performance is better then advertisers will spend more money and ultimately generate more revenue for Google. And with most of Google’s decisions, somewhere along the line if you trace it back it’ll link back to revenue.
So, what does it mean for people working in media? Well it’s probably a scary thought handing over all of the control of your campaigns to algorithms and automated processes. When, essentially, it’s a performance agency, when the buck stops with us, that can be quite a scary thought. And you might want to control all of your bids and all of your targeting manually, and you might think that you can outperform the algorithm. But the way I see it, there’s only one way that the future’s going really.
I mean we’ve seen it with Google in the past trying to remove manual options, things like ad rotation using terminology in the interface. So, for the optimised version of ad rotation, the label is the ideal setting for most advertisers. Without any optimisation, if you set it to rotate indefinitely, not recommended for most advertisers and you get a warning. Little things like that. And they’re not afraid of making big changes and ultimately turning things off.
We’ve seen with the new AdWords interface that was slowly released over 2017. It received a lot of negative feedback initially and people have continuously been using the old interface and not wanting to switch over. But they’ve used little tips and tactics to try and get people to move over, making it the default interface when you log in, only releasing new products and features in the new interface. It’s been suggested that the old interface will be sun-setted in Q1 of this year. I don’t think that’s going to happen, but they communicate in that way obviously to try and move people over. They’re smart, they know what they’re doing.
So, basically, I don’t think we have a choice, that’s the way the future’s going. And if you want to stay in the game and still be relevant in a few years then I think you’re going to have to embrace it.
So how do we do that? I think, firstly, when using automation and machine learning algorithms it’s important to understand what’s going on, what the algorithm’s doing, what data it’s using, how it’s making its decisions, and essentially allowing us to better understand what the outcome was and why that happened. And, if performance was good or bad, why was it good or bad? Not just sort of taking that and going: “Well it didn’t work for me”.
Granted, fully understanding how some of these products work isn’t always easy, but Google tends to be pretty open with the data that they’re using to run these algorithms and how we can affect the outcome. Ways we can affect the outcome? So, things like adjusting settings or feeding in richer data sets. For example, if you’re using bidding strategies that are adjusting bids based on audiences that you have in AdWords or DoubleClick, making sure you have lots of audiences set up, useful audiences of bucketed users that it’ll be able to analyse and make decisions on. Essentially, giving it every chance of winning.
And lastly, don’t be afraid to test things. I know we’ve tested a lot of stuff internally. You need to be scientific with you approach. Run tests, analyse the outcome. If you don’t get what you wanted, try to understand why and not just give up and say: “Well that didn’t work for this client” without fully investigating. And be resilient.
And ultimately, if we get it right and the algorithms perform well, it’s a much more efficient way of managing campaigns, obviously. If we can remove the need to analyse rows and rows of data manually, we free up a lot of time to start focusing on other things, things like strategy, planning, creative and user experience.
And bringing it back to us, so from an agency point of view, I think that’s very exciting. Our position as a data driven performance media agency means that we’ve been at the forefront of this for a while. But we’re all on this path of maturity together when it comes to advertisement technology, learning how to best utilise advances in machine learning and AI. And I think we’re in a pretty privileged position as well, working in marketing. Automation can help us do our jobs better, rather than replacing us completely like in some other industries.
So, thinking about that. What’s going to differentiate you, whether you’re a marketer or a brand, or work in a media at an agency, when automation’s dominating many of the activities that we carry out today it’s going to be back to the old school. Focusing on core marketing skills like strategy, planning, creative, providing insight. And in that landscape, we need to be better whole brain marketers. Sure, you’re still going to need technical skills but we’ll have more time to focus on the creative side of marketing. That’s my prediction anyway.