Machine learning will be more than a marketing buzzword in 2018. It’s already powering large chunks of the platforms you use to reach your target audiences (Google Search, AdWords, etc.) and the technology is only going to play a bigger role in the way people use the web.

Machine learning isn’t reserved for the tech giants, though. Marketers can also use the technology to unlock the full power of big data, optimise their campaigns like never before and automate a wide range of tasks to get results faster.

Google knows this better than anyone and it’s starting 2018 with some advice for advertisers about using machine learning to make big things happen this year. You’ll find the three-part guide called Grow your business faster with machine learning over at the AdWords Blog but we’ll be covering the same points and expanding upon them in this article.


Machine learning is all about solving problems

As Google says, machine learning is all about solving problems:


“Rather than spending hundreds of hours manually coding computers to answer specific questions, we can save time by teaching them to learn on their own. To do that, we give the computer examples until it starts to learn from them–identifying patterns, like the difference between a cat and a dog.”


So what kind of problems can machine learning solve? Well, it’s the technology responsible for putting driverless cars on our roads, predicting illnesses before they strike and translating conversations in real-time. So it’s already preventing road deaths, saving lives and breaking down languages barriers – among other things.

For advertisers, though, machine learning is there to solve a very different set of problems.

The first problem machine learning solves for marketers is the challenge of utilising big data. As much as we’ve been talking about big data in the marketing industry for years, it’s only the likes of Google, IBM and the other tech giants that have been able to utilise it. Processing large volumes of data takes time and huge teams of data scientists, which rules out the vast majority of businesses and advertisers in the big data race.

Machine learning changes all of this.

With intelligent algorithms capable of processing almost infinite amounts of data in a short space of time, even the smallest business can unlock the full potential of big data. This brings a new world of opportunity for advertisers across every industry.


Machine learning as a growth accelerator

One of the most important aspects of machine learning is it makes tasks faster to complete and their results more reliable. Here’s the example Google has to offer:


“We know that choosing where your ads show and manually adjusting bids is time consuming, leaving less time for strategic tasks, like capturing the latest trends or entering new markets. Google’s machine learning considers billions of consumer data points everyday, from color and tone preference on mobile screens, to purchase history, device and location. With products like Universal App Campaigns and Smart Bidding, it’s now possible to use this data to help deliver millions of ads customized for your customers, and set the right bid for each of those ads–in real time.”


Of course, Google wants you to use features like Smart Bidding and its other automated, machine learning tools. These can be very useful, too, but you’re not limited to the machine learning systems Google or other advertising platforms provide.

You can create machine learning innovations for yourself.

For example, you can build an algorithm based on your own historical data (and no-one else’s) that analyses bid performance over the past five years to automatically optimise your bids throughout the day. This way you know exactly what data is being used and what the objective is without that nagging feeling Google is more worried about maximising clicks for its own profit rather than overall performance for yourself.

That’s not all, either.

The learning aspect of machine learning is important. Essentially, this works by cross-referencing dataset, looking for patterns and then applying these to future calculations. This is how a driverless car learns when to break and when to accelerate; it’s applying previous calculations to entirely new scenarios.

So you could take the bid automation algorithm we talked about earlier and then incorporate a new variable that considered annual holidays: Christmas, Black Friday, Valentine’s Day, etc. Then you can add another that considers the time of year and climate to cater for differences between the summer, autumn, winter and spring. And why not go a step further by tapping into local weather resources to calculate the impact of today’s weather on the performance of your ads?

Eventually, your data insights will be rich enough to know that your target audience is most likely to search for holiday packages on rainy Monday mornings during February and March after the Christmas expenses are long forgotten. Or that your ads perform best on Friday evenings after people clock off from work for the weekend, except for the final weekend of the month when most people get paid and hit the pub instead of Google.


Creating a more powerful AdWords with machine learning

Something we’ll see more of over the coming years is new machine learning features work their way into the AdWords experience. We’ve already mentioned Smart Bidding but you can also see this in Dynamic Search Ads, which automatically fill in keyword gaps, and Smart display campaigns, which automatically create ads, target audiences and set bids to maximise performance – plus a bunch of other AdWords features.

Again, you don’t have to rely on Google to implement its own version of machine learning in AdWords, though.

Let’s say you’re looking to reduce ad spend and you decide to pause all campaigns that are spending more than £100 per month but haven’t converted in the last 30 days. To do this manually, you’ll need to first segment all of your campaigns by monthly ad spend, filter out all of the campaigns that have converted in the last 30 days and then manually pause each campaign that’s remaining.

This can be a mammoth task if you’ve got a lot of campaigns running. Worse still, you might find out you haven’t cut ad spend enough or you’ve cut profit too much and have to start all over again.

With machine learning, you can create an algorithm that automatically pauses all campaigns based on the criteria you set. Crucially, you can change these criteria in one place and apply them to every relevant campaign as you choose, without manually ploughing through different campaigns.

You can take this a step further, too, and create a notification system that warns you when campaigns haven’t converted for 20 days and group them based on similar characteristics, allowing you to test changes at scale and improve performance before it even becomes an issue.


Making sense of today’s consumer journey

As Google puts it, “The days of predictable web sessions are over, replaced by bursts of digital activity throughout the day on multiple devices.” Which means advertisers need to consider a wide range of signals that impact campaign and ad performance: device, time of day, location, user intent and all kinds of other factors.

As the consumer journey becomes more complex it also becomes more difficult to predict and this is where machine learning will make one of its biggest impacts on advertising.

Of course, Google wants you to use features like Smart Bidding to cater for these more complex consumer journeys but this will never unlock the full power of machine learning for you. With tools like Google’s Cloud AutoML and Microsoft Azure you can build and train your own machine learning models using drag-and-drop interfaces, without any code.


Source: Azure


You don’t need to rely on the machine learning tools and features built into platforms like AdWords. You can build your own, using your own data and designed specifically to your own marketing goals. For example, you can build a system that maps user actions across all of your advertising touch points (Google Search Network, Google Display Network, Facebook, etc.) to build a more accurate picture of how people interact with your brand across the entire consumer journey and how your ads perform as a single strategy.

These kinds of insights are what make machine learning so exciting for advertisers – large volumes of data that can be processed almost instantly with an automated system.


The aim of this article was to highlight the key areas where Google thinks advertisers should be using machine learning in 2018. This is important because it gives us an idea of where Google will take AdWords over the next year or so but we also want to make it clear that there’s more to machine learning for advertisers than using features like Smart Bidding – not that there’s anything wrong with AdWords’ automated bidding features when they’re used in the right scenario.

The real magic of machine learning, though, is taking control of automation for yourself and building insights from your own data. And the latest breed of machine learning tools is making this technology accessible to almost anyone. You don’t need to know how to build these algorithms, you just need to know how you can use them to make great things happen.