Machine Learning and Digital Advertising
Understanding machine learning is key to producing better results in digital advertising. Machine Learning is an area of study that has been around since the 1950s, but it's only recently become more accessible to a wider audience. In this article, we will explore machine learning from the perspective of a layperson and discuss some ways that you can harness this technology in your business.
Better Modeling, Better Predictions
Machine learning was invented to find better models for understanding data. It is basically a machine that can learn from data and make conclusions. The machine does not have to be told what the right answer is, it will find patterns in large datasets without any pre-programmed instructions about how to categorize or sort information. As machine learning takes more guesses at predicted outcomes, it can become more powerful and accurate. That being said, we have to understand that the machine will not be able to predict outcomes with 100% accuracy. The ultimate goal of machine learning isn’t even to be 100% accurate, it is primarily about reducing the gap between expected outcomes and actual outcomes.
In marketing automation, for example, machine learning allows you to generate customer journeys through your funnel that connect with users on a more personal level. Google's audience groupings are utilizing machine learning in just this way. Google collects data on user interests and search patterns. It then groups people into potential groups to try and serve more relevant content and ads. The machine that Google is building accepts that it may not serve the right content at all times, but it continually works to make better predictions with each and every interaction on the internet.
The Learning Phase of Automated Bidding Strategies
In both Google Ads and Facebook Ads, there is a period of learning where the machine is building up its model of data before running at peak performance. The learning phase of automated bidding strategies means that the machine may make some mistakes in predicting which ads will perform and which ones won't. So, if you want to test out a new strategy for targeting users on Facebook or Google Ads, it's best to start with a small budget before ramping up your spending as machine learning processes more data over time.
Additionally, we shouldn't be afraid of the learning phase. Many marketers hesitate to make changes to ads because they might send a campaign back into the learning phase. But machine learning is so powerful because it can continuously learn. The potential for machine learning to do tedious work while humans do more meaningful work is huge.
The period of time spent in this phase depends on how much data you want the machine to process before running at peak performance. It may take days or weeks for some machines and others could be just a few hours. If you are not testing, then machine learning will not have enough information about your campaign. So keep testing creative, location targeting, budgets, and anything you can think of to improve performance. The key is to pull back on changes when you realize what is working best. Always ask yourself, what does peak performance look like for this campaign?
How to Work with Machine Learning
What are some of the best ways to tell a machine what to optimize for?
Measure What Matters Most
For one, we need to set our conversion strategy on only measuring conversions that are valuable to our business. All other vanity metrics should be left out of the conversion column.
For lead generation, for example, we should set the conversion goal triggers not just on lead form submissions or phone calls, but instead on leads that turn into paying customers in our CRMs. In this way, Google can figure out the weighted value of conversions. Leads are great, but leads that turn into revenue are better. Google Ads can attempt to get both in different volumes so that you can remarket to non-revenue generating leads and get a better return on ad spend in the future.
Check out Hubspot's article on understanding what marketing automation is.
Remove as Many Restrictions as Possible
One thing that makes all marketers uncomfortable is the idea of giving up manual controls for automation. But what will really unleash the optimization power of machine learning is a loosening of restrictions in regards to budget, placements, creatives, and audiences. Eventually, we will learn when and where to trust machine learning and how far to push the algorithms. Until then, it would be wise to slowly test removing one restriction at a time and measuring performance over long enough periods to make solid decisions about removing the guard rails.
Use The Power of Google's RSLAs
RSLA stands for Remarketing Lists in Search Ads. Remarketing lists enable you to tailor your search marketing campaigns with potential customers who have previously visited your website. RSLAs are a powerful tool that leverages Google's machine learning to get your ads in front of browsers that have already interacted with your website. Remarketing ads are 70% more likely to convert compared to ads that a user has seen for the first time.
There are two ways to implement remarketing lists with search ads.
You can optimize your bids by implementing a remarketing strategy that targets visitors who have interacted with your website in the past. For example, you could increase your bid for those who visited your site within the last 30 days or show a different ad to people who place items in their cart but fail to purchase them.
A strategy for savvy site owners is to increase your conversion rate by bidding on additional keywords while targeting users who have previously landed on your website. This way, you are optimizing your bids based on the behavior of the visitors of your site instead of people from any other location that may be less likely to convert.
How to Keep Up With the Pace of Automation Technology
The simple answer is you can't. Sorry to break it to you. I hate being the bearer of bad news. Automation technology is changing too quickly to know everything about it.
However, there are some things that can definitely help novices like myself and many other marketers. For one, follow Google and Facebook closely on their insider training sessions. You can get access to these by being a Google Premier Partner or Facebook Business Partner. Many of these training are surface level, but they often give major hints on how to adapt to the increasing power of automation. In some cases, they tell you exactly what to do to become a top performer in your market. You just have to attend them with an open mind to vet what is actionable and what is fluff.
Another thing that you can do is read machine learning blogs. These can be a major resource for staying up-to-date with machine learning and how it will impact the future of our industry.
In Summary
Machine learning is a powerful tool that can do things humans simply cannot, but it’s not perfect. The best way to use machine learning in your digital marketing strategy is by testing and tweaking what the algorithm spits out. Automation technologies are moving at an astonishing rate these days, so keeping up with them may be challenging. Luckily though, there's automated technology for everything from grocery shopping to handling notifications on our phones!
Follow my blog posts for more information about how you can apply automation or machine learning principles in your own business or life.
