How Machine Learning Drives Personalization at Scale in Adobe Target
Using machine learning and artificial intelligence, Adobe Target delivers personalized experiences to your customers in a matter of milliseconds.
Personalization isn’t just good for your customer experience — it drives business impact.
Organizations that personalize customer experiences reduce acquisition costs by 50 percent, increase revenues by 15 percent, and optimize marketing spend by 10-30 percent.
You know your customers want and expect personalization, but trying to create this experience manually within your marketing technology stack doesn’t scale.
To deliver highly personalized experiences, you need to automate your process, and that’s where artificial intelligence (AI) and machine learning can be really powerful. Machine-learning algorithms allow you to automatically identify relationships between data and learn the best experiences to show based on that data, without having to be explicitly programmed.
This advanced technology — and the right data — can help your marketing organization build better models that drive personalization. Adobe Sensei, the AI and machine-learning technology that powers many of Adobe’s products, brings machine learning-based personalization to Adobe Target, the optimization engine within Adobe Experience Cloud. Thanks to features like Auto Target and Automated Personalization, Target does a lot of behind-the-scenes work to ensure you’re using advanced machine-learning technology — all with just the click of a button. Here’s how:
Marketing decision-making made easier
Auto Target and Automated Personalization are two of the most powerful AI-driven features in Adobe Target.
Based on an individual customer profile, Automated Personalization shows a specific visitor different sets of offers or messages to increase the chances that a customer will convert. With Auto Target, personalization is based on a set of experiences defined by the marketer, and the delivery of these experiences is automated thanks to machine learning. Auto Target and Automated Personalization are both based on the same powerful AI algorithms.
The main difference between Automated Personalization and Auto Target is that Automated Personalization helps you understand what the lift and impact of your marketing activity is at the offer level (for example, a holiday discount offered to a specific customer). However, Auto Target allows you to understand the impact of your marketing activity at the experience level, such as a lead nurture campaign or a multipage journey on your website.
Applying AI to your real-world scenario
How does all this behind-the-scenes technology work when a real customer engages with your brand?
Let’s say you’re a travel company, and you want to deliver a personalized offer for a tropical beach vacation to Sarah, one of your customers. Adobe Target collects thousands of data points or attributes about your customers, including Sarah — where she lives, where she is now, which browser she uses, her favorite color, what the weather is like where she is, and more.
These attributes also can be related to site behavior (new or returning visitor), temporal variations (time and day of week), channels (mobile and desktop), referral source ( social media ad), and even your own internal CRM data (whether Sarah is a loyalty club member, for example).
That said, you don’t need all of this information to effectively target Sarah. If you’re selling a vacation, for example, it doesn’t really matter that Sarah’s favorite color is blue.
Within Adobe Target, Adobe Sensei gathers the information that makes the most sense for your offer (and tosses out random outliers) to build a machine-learning model that can find patterns in the data that is most relevant to your goal (to get Sarah to convert), and better target each of your customers. In this example, for simplicity’s sake, we’ll say the model decides to use three features about Sarah: she’s 35, lives in Boston, and uses Google Chrome. Age, current city, and browser are each called a “feature.”
With Adobe Target, each individual visitor — Sarah included — gets a visitor profile populated with these features. At any time, any combination of features can be used to craft a personalized message, offer, or experience.
As you begin personalization, these features form the backbone of the models to be tested. With machine learning, Adobe Target is able to automate personalization and A/B testing, which means the models get even more powerful over time as they are fed more information about final outcomes.
Delivering the wow
Before reaching that point, though, your travel company must figure out which content Adobe Target should consider to deliver to Sarah right now. To do this, Adobe Target takes a snapshot of Sarah’s visitor profile, transforms the data, then decides which features get included in the machine-learning model. The selection can depend on a number of factors, such as which content has worked in the past or whether you’re trying out a new offer.
Once the model is built, Automated Personalization and Auto Target powered by Adobe Sensei use machine learning to assemble and deliver a personalized experience for Sarah. The model decides in real time which content and offers to show Sarah that will be most persuasive.
But here’s the catch: Don’t shelve potential — but unused — content too quickly. After all, Sarah might not have jumped at your 30 percent-off coupon or your voice assistant campaign for a number of reasons, some entirely unrelated to the nature of the offer itself.
This problem is what is called a multi-arm bandit situation: You might hate the long grocery line you’re stuck in, but if you move to that other more tempting one, will the line really move faster? You won’t know unless you’ve tested it. The way around this is the multi-arm bandit approach, used by Adobe Target, which essentially balances between testing new options versus exploiting what option looks like the best today.
Test, analyze, repeat
So how does this all factor into the personalized experience that Adobe Target delivers to customers like Sarah? When she visits your site, Adobe Target quickly scores the personalization models and then determines how it will feed content based on a multi-arm bandit approach.
The best part? Adobe Target makes all these decisions within milliseconds, so you don’t lose an impatient customer.
Back to Sarah, who, at this stage, gets one of two offers: “Book your tropical getaway, save 30 percent” or “Tired of Boston winters? Take 30 percent off a tropical getaway.” Will she bite? Whether she does or not, you will want to analyze the results of your personalization. If Sarah bought your vacation, what sold her on it?
Equally important — if Sarah didn’t book, what happened? And what can be changed in the future?
Traditional post-mortem analysis with machine-learning systems is tricky because of the dizzying array of data involved. Advanced AI algorithms find so many different patterns in data that they leverage to make decisions that it’s known as the “black box” — you know what went in, but you have no idea what patterns were most important to arrive at the solutions that are spit out.
Adobe Target, however, delivers personalization insights reports, which allow you to make educated decisions about whether your campaign hit a home run or struck out spectacularly, and then fine-tune future campaigns accordingly.
Customers like Sarah want you to deliver personalized experiences. When Adobe Target uses machine learning and artificial intelligence, they get just that. And Sarah gets to avoid Boston’s nasty winter while enjoying the beach, and you get the best ROI for your marketing campaign. That’s (almost) as good as a vacation.
Data science is all about trade-offs
It’s a given that your customers want and expect personalization. But it’s also a reality that you don’t have unlimited resources.
Understanding the trade-offs and the role they play in your business goals is critical to success in your personalization program — and how you use machine learning more broadly. Focus on better data and better models, and your personalization will also be better.
Do you have thousands of “Sarahs” to reach with a personalized brand experience? By tapping into the power of Adobe Sensei, Adobe Target makes it easy to repeat this process for every audience segment, every time. Read more about how Adobe is using artificial intelligence to help automate and scale your marketing processes.