Most marketers believe in the benefits of personalization. They agree that their customers expect personalization and that if they delivered more relevant experiences, they would increase their bottom line.
Yet many people I speak with are still providing static experiences to their prospects and customers. The most common responses I receive when I ask why are: (i) the majority of their traffic is anonymous and therefore personalization won’t be effective and (ii) they do not have the necessary resources to dedicate to personalization. These reasons may have been valid in the past, but they no longer hold up in the age of machine learning.
In this blog post, I will explain how you can use everything you know about each website visitor, even if they are first-time and/or anonymous visitors, to automatically select the optimal experience in the most efficient manner possible.
There is No Such Thing as a “Zero Data” Person
When an anonymous, first-time visitor lands on your website, the welcoming experience can and should be more relevant and effective than the default experience.
A lot can be identified about a person even if you don’t know his identity. Here are a few contextual data points that can be recognized immediately upon a person’s arrival to your site:
- Browser, device type and operating system
- Acquisition source and referring domain
- Geographic location (country, city, state)
- Weather in that location
- Time in that location
- Day of the week
- (Optional) Third-party data such as demographics, firmographics, in-market shopping behavior, derived affinities/interests, and more that can be provided by a DMP
Wouldn’t this be helpful information for you to use to decide in real time — the second a person arrives to your site — whether to feature sandals or rain gear, vacation or business travel, credit cards or auto insurance?
And that’s just the data you can capture right away. If you’re using Evergage, the moment your visitor starts browsing, more information is being captured, stored and analyzed in real time at the per person level. This data includes things like:
- Products/pages/content viewed
- Promotions engaged with or ignored
- Segment membership
- Affinity towards your business context (such as product categories, blog topics, etc.)
- Active time spent browsing in categories, brands, colors, styles, topics, etc.
- Predictive scores
This type of browsing data is powerful and should factor into which assortment of products, homepage messaging, imagery or content someone should see on your site going forward.
Every piece of real estate on your website holds an opportunity to engage your customers — to help them find exactly what they’re looking for in a sea of options or to discover new and relevant products...all while a person is still anonymous to you.
Thus, even if the majority of your traffic is anonymous, there is still so much that you can know and use about each visitor. If this contextual and browsing behavior is used to help power your welcoming experiences (plus experiences going forward as you learn more), it will lead to a better customer experience and better results.
In order to optimize the effectiveness of this data, you need to have a system that enables you to make real-time decisions at scale that is continuously learning and optimizing.
Time to Shift Your Thinking About the Effort Required
Another common reason companies still aren’t personalizing is because they believe they do not have the necessary resources to dedicate to personalization — even if they believe in the idea.
They may feel this way because when many marketers think about making use of their visitor data for personalization, they only think about rule-based personalization. With this type of personalization, marketers manually select an experience to deliver to a specific group of visitors (such as showing people in New York City urban imagery or urban product trends). This is certainly a valuable way to create relevance. It could be a good way to dip your toes into personalization — selecting a few predefined segments to speak to. But when you are trying to scale to become individually relevant to each person, there are a few challenges to overcome with a rule-based approach.
First, marketers need to manually create many segments, plus the experiences that correspond to those segments. This process takes up a lot of dedicated resources and can get confusing and hard to manage if you are looking to expand beyond a few core segments.
Second, there’s still a good amount of guesswork involved. How do you know which rules to make? Are you utilizing all the data available most effectively? Which segments do you target? How do you know that the experiences you spent time building actually produce the most optimal experience for each person in that segment?
A rule-based approach is useful, but if you are investing in a solution that only provides this capability or you only use this capability, you aren’t maximizing long-term ROI, because you can’t achieve your dream of offering unique experiences optimized to every individual.
It’s time to shift your thinking away from relying on preset rules and start using machine-learning models. It’s the only way to scale.
Bringing it All Together with Machine Learning
As described above, even if the majority of your website traffic is anonymous, you still know a lot about them. You can certainly leverage rules and segments to act on this data to deliver a more relevant experience to anonymous visitors. But machine learning can make the process more efficient and effective.
Imagine having ten variations of a homepage hero banner (various imagery, promotional content, product promotions, etc). If you wanted to make use of each one using manual segmentation, you would have to define ten or more segments and match up each one with what you think is the most relevant to each segment.
If instead you let an algorithm like Evergage’s Contextual Bandit make the decision for you, all you’d have to do is design the experiences and allow the algorithm to decide which one is optimal for each person based on all the available data — even for a first-time visitor. In addition, Contextual Bandit is always learning to improve its effectiveness — without you needing to take any manual action — and it even factors in the business value of each experience to your company.
Personalization doesn’t have to be hard. You want to utilize all the available data, even for anonymous visitors, and make the shift towards machine learning-driven decisioning. It will save you time and resources, eliminate guesswork, help you uncover insights and unlock new opportunities in your mission to provide relevant experiences at the 1-to-1 level while driving more engagement and conversions.