In order for AI to work, precision is paramount - marketers need to focus on data quality, underpinned by smart technology to deliver accurate/ exacting/ faultless customer journeys.
When people think of AI, they get pictures of grandeur - fancy robots and machines that perform amazing tasks for humankind. However, in the marketing world, AI is actually another type of incredible - it is driven by machine learning algorithms that discover patterns in big data.
As ironic as it sounds from a human standpoint, AI is the next step and key to unlocking the full potential of personalization. Although we, as humans, have the capability and knowledge to create detailed, personalized customer journeys, the possibilities are endless when it comes to the potential paths a customer can take from start to finish. In order to help us manage these limitless customer journey paths, AI is needed to streamline the process and enable us to free up our resources so we can focus on the bigger picture of serving our customers with the best experiences possible.
We sat down with Relay42’s resident Machine Learning Expert Duco van Rossem to get his insights on AI, digital marketing, and the future.
There’s a lot of data transmission going on - some clients generate over 2 gigabytes of data per day. Within that transmitted data is a lot of information. Because of the number of variables and factors in play, machine learning can significantly assist humans in marketing. Humans are very inefficient at processing data, even with a whole team at play. While we tend to get more overwhelmed with more data, machine learning algorithms actually work better with more data; so the more data you have, the more complex techniques you can use (i.e. deep learning), the better outcomes you get with machine learning. The algorithms can pick up on very complex trends and patterns while humans can only really think in cause and effect terms, with only 2 to 3 variables at a time.
When it comes to personalization, because there are so many factors in serving the right message, to the right person, in the right context, it doesn’t make sense to take a hands-on approach with handcrafted rules for every conceivable preference someone could have. Machine learning can support marketers in delivering on the promise of personalization at scale.
Not only can AI help marketers to deliver on personalized customer journeys, but it can also help with reducing ad spend as well. AI can automatically decide to target customers who are less likely to buy, or book, on an owned channel rather than paid. It can also target ‘higher value’ customers on paid channels, like display and push messages based on urgency, such as the time left to book a flight in the case of airlines.
The main challenge that we have with machine learning is getting clean, structured, standardized data. You can’t just plug raw data into an algorithm - you have to really think about both how you structure your data and how to make it fast enough to predict things on a large scale, in real-time. People usually expect that you can just plug data into an algorithm and everything works out by itself, but there’s actually a lot of nuances and context for your problem that really matter when you want to use your algorithms effectively. We need an effective way of going from raw, unstructured data to something that is clean and ready to use in machine learning. That is something many people in AI are working on - to scale this approach.
Generally, in machine learning, there is lots of research done with images and text, and those have become standard problems to solve. For example, a lot is known about dealing with image data, but dealing with customer journeys is a less standard problem. There is much promise for machine learning in this space of personalized marketing, where it holds the ultimate key.
It’s more a question of the rate of progress. There’s always going to be progress, but in some areas, it will be faster than others. The possibilities for AI depend on what an acceptable level of accuracy is for an algorithm so that it can be used in real life. The more work done on improving algorithms, the more potential real-world applications. At Relay42, we believe in practical implementation of AI. While AI is useful in automating customer journeys, we do need to be realistic about how it’s applied. With quality algorithms and smart marketers working hand-in-hand, this is the solution to the complicated issue of optimization when it comes to personalized customer journeys.
Want additional insight on AI? Read The Drum's article: How AI can maximise impact across customer journeys