The future of online advertising
In recent years, major advertising platforms such as Google Ads and Facebook Ads have made big efforts to improve their Machine Learning technologies. As a result, their Machine Learning is getting better and better and needs less data. The consequence for online marketers is that much work, such as setting bids and target group targeting, is fully automated by these platforms. How will this continue in the future? The pace of automation is expected to continue at the same pace, with Machine Learning playing an (even) more important role. Responding to this as an online marketer will become critical to future success. This article explains the limitations of Machine Learning and how to get the most out of it as an online marketer.
How Machine Learning Works
Machine learning is a self-learning system. Major ad platforms such as Google Ads and Facebook Ads use it for the following, among others:
- Setting up bids
- Showing, and compiling, the most relevant ad
- Target group targeting of campaigns
For all of these components, Machine Learning looks at data from the past to make a prediction about the future based on that data. In practice, for example, this means that for each person searching, Google Ads’ Machine Learning will calculate for you the expected conversion rate and then determine the bid. Machine Learning on ad platforms provides better results and significant time savings.
The limitations of Machine Learning
In addition to all the benefits of Machine Learning, there are some limitations that you need to be aware of and consider when deciding whether and how to deploy it. The main limitations are as follows:
- Amount of data: the more data available the better it works. For Google Ads campaigns running on a Target ROAS bidding strategy, it is important that you have enough transactions (at least 30 to 50 per month) to get good and stable results.
- It steers for your input: Machine Learning does not contain intelligence like we as humans do, it is not self-thinking. Which means that machine learning only controls on the input we give, if this is not good then the output is often not good either.
- Sudden events: Machine Learning benefits from constancy. Drastic changes in external factors such as weather are variables not taken into account.
The next section of this article will focus on how to best use Machine Learning and how you can make a difference over other online marketers.
Deploying Machine Learning Optimally
To maximize Machine Learning from ad platforms, it is important to focus entirely on the input you provide. In addition to important factors such as your website and proposition, within ad platforms it is important to focus primarily on the following inputs:
Objective:
On what objective you are driving Machine Learning.- Creation: The texts, images and videos you make available.
Own data:
Sharing proprietary customer data.
Objective:
The objective of your campaign should be central to how you drive bidding strategies and campaigns within ad platforms. In doing so, always focus on quality rather than quantity. Below are some examples of how to do this optimally:
- B2B: Link the value and quality of leads back to Google Ads or Facebook Ads via
offline conversion imports
.
Ecommerce:
Correct your sales on returns and/or import “margin” as a conversion action within Google Ads and Facebook Ads.
Omni-channel:
Drive not only online revenue but also optimize for store visits.
The more qualitative the conversion actions you pass along to Machine Learning, the better the output will be. So as a B2B party, do you want no more whitepaper downloads from students? Use offline conversion imports to return the value of a lead.
Creation
Advertising platforms are not yet at the point where ads are fully automated. Machine Learning does get better at composing an ad to show the ideal message at the right time. The better the creation used the better the result. Therefore, the following is important:
- Text: Combine as much relevant text as possible, see what your competitors are doing, and hire a good copywriter whenever possible.
- Images: try to be creative and provide very quality images. As an example, the image above that stands out extra because the product is just a little twisted.
- Videos: Create distinctive videos.
When creating, try to test as much as possible what works and what doesn’t. Based on that, you can continue to improve the creation and feed Machine Learning with the best videos, visuals and texts to the max.
Own Data
Advertising platforms have a lot of data from users. This data is used to target, for example, certain interests. Data they don’t have, however, is your customer data. The better ad platforms know your customers the better Machine Learning can help in audience targeting. Sharing customer data is becoming increasingly important as the shelf life of cookies recedes and becomes virtually impossible to use in the future. The advice is to share the following with ad platforms:
- Customer lists: Upload automated customer lists in Google Ads and Facebook Ads, based at least on email addresses and phone numbers.
This can be done through the Google Ads or Facebook Ads API or with integration tools such as
Zapier
. Try to deliver the customer data in as many relevant segmentations as possible to give Machine Learning maximum input for audience targeting.
Conclusion the future of online advertising
In recent years, we have seen ad platforms automate a lot; this is expected to continue in the coming years. Above all, go with this automation and take advantage of the power of Machine Learning, but also be aware of its limitations. The most successful online marketers of the next few years will increasingly focus on how they drive Machine Learning. Are you already driving Machine Learning the right way? This article has given you the tools to use it to its fullest potential.
