/ Product

How we use predictive segmentation to convert visitors into buyers

Data, data everywhere. As marketers, we’re used to collecting furious volumes of data for a sole purpose - to take a look into a complexity of consumer behavior and translate it into an intelligent customer interaction. And as consumers, we’ve already felt it:

  • Google autocompletes your search queries
  • Facebook news feeds are tailored based on your “likes” 
  • Tesla is applying a mesh network of AI techniques where one car helps all other cars in the network learn on-the-fly

It’s no secret that harnessing large volumes of data quickly and accurately will positively affect your revenue stream through acquisition, conversion, and retention. The trick is - the rate of consumer data and intent often vastly outpaces our current technological ability to act on it.

Marketing cloud companies have released a slew of complex products that attempt to allow marketers to harness this vast data volume. However, most technology stacks struggle to effectively enable online marketing professionals to mobilize their marketing at the speed that consumers are researching and conducting their online decisions. It becomes hard to keep up.

The result of data that isn’t continuously modeled is a significantly missed opportunity. It’s a failure to connect with modern consumers in the way they expect (and demand).

Innovations are training consumers to expect more from online businesses, creating a new standard where marketers must not only understand consumer needs, but anticipate them.

But don’t reach for your crystal ball just yet - by implementing predictive segmentation and notifications, you will convert your visitors into customers by anticipating their needs.

Predictive Segmentation to the Rescue

Predictive segmentation helps you to notify your visitors about any relevant offers or actions they should take, based on their individual consumer fingerprint. It becomes possible due to the machine learning and AI, which are confidently dominating B2C verticals.

The AI-enabled marketer armored with predictive segments will be able to:

  • Leverage smart modeling to predict each consumer’s likelihood to perform any action;
  • Specify a set of personalized possible notifications to engage each consumer;
  • Automatically adapt the journey for each individual consumer, along a predefined funnel;
  • Deliver the best next product, content, or offer - every time.
  • Send notifications at the right time, when a consumer is likely to engage, and when a brand needs to.

How can you implement predictive segmentation and personalized notifications?

Traditionally, marketers have lumped audiences into broad groups based on attributes like location or simple product category based intent. To benefit from a smarter approach, you should segment your visitors into different buckets, as narrowly defined as possible. An effective way of doing this is to use behavioral segmentation, i.e. putting your users into different segments based on their activity over one or more web sessions.

Micro-targeting requires the marketer to know their audience niches and their needs. This isn’t always the case nor realistic when you’re dealing with hundreds of products and categories, and with different consideration periods prior to purchase. Can a predictive approach help?

Online behavior can be shifted towards increased conversation rates by applying predictive segmentation to personalize shopping journeys. Let’s take a look at an example of how online subscription company Barre3 could acquire subscribers by using our techniques.

Use Case: Barre3

Barre3 created a healthy lifestyle hub - apart from having studios in 30 of the United States, Canada and the Philippines, they offer workout accessories, apparel and online courses. The last comes in the form of a monthly / yearly subscription.

B3_homepage-1.png

Barre3 Homepage

By subscribing, members have access to exclusive workouts, nutritional guidance and expert support. Barre3 also has separate domains for its blog and online store.

B3_blog-1.png

 Barre3 Blog

What If?

Let’s explore an idea of how Barre3 could acquire new subscribers. Their bounce rate is around 41.6%, and an average visit lasts 2 minutes, with 2-3 pages per visit on average. However, their subdomain - blog.barre3.com - is responsible for 14% of all visitors, which is remarkable, considering that shop.barre3.com, for example, is responsible for roughly 5% of visitors. Since the total visits to the website are about 612,000 per month, we’re dealing with nearly 86,000 blog readers who could become subscribers, if only given proper guidance through the funnel.

We are looking to create a campaign that will leverage the power of existing blog content and prevent Barre3 from any wasted ad spend - according to public data, we see that the majority of Barre3’s display advertising efforts are aiming at the “Subscription” and “Store” pages.

Let’s create a campaign that will encourage visitors to become subscribers and prevent losing money on advertising. Here’s what we do:

  • Link domains. First, let’s identify the shared visitors of our two domains,  barre3.comand blog.barre3.com. Using a platform like Intempt, we specify a tracker for 2 domains to understand visitor behavior across domains and leverage the data in the future campaign. Now, we store every interaction by a visitor on both websites to understand their behavior.

B3_2domains.png

Linking two domains in the tracker

  • Identify a drop-off point. We’ve assumed that visitors leave the blog after viewing 3 pages. To back our hypothesis with data, we slice two narrowly defined segments out of the funnel and compare how many visitors have fallen into each segment. We create a Segment 1 for visitors who browse 3 blog pages (behavioral property Count Of) and a Segment 2 for visitors who browse 4 blog pages (behavioral property Count Of). Here’s what it could look like:

comparing_segments-1.png

Segment 1 audience vs Segment 2 audience (marked in red)

Based on the gap we noticed between the two segments, we know a visitor is likely to leave, so we prevent the drop-off with a targeted notification.

  • Segment our visitors. We specify a target group by blending past behavior (fact) and future behavior (predicted). For this segment, we are focusing on first-time or returning visitors who haven’t subscribed (behavioral property Has Not Done), browsed 3 or more pages of the blog (behavioral property Count Of) ,and are predicted to drop off without subscribing (predictive property Will Not Do.)
  • Time to campaign! For this campaign, when 3 blog pages are reviewed, we know the possibility of a drop-off is high based on the average number of pages per visit. It’s a good time to send a notification inviting the visitor to check out barre3 subscription. Final look:

B3_notif_final-1.png

Simulated Scenario - Predictive Notification

B3_notif_final_zoomed-1.png

Zooming In

Why This Notification At This Time?

  • The visitor has engaged with at least 3 blog posts;
  • The visitor is new to the subscription model;
  • The visitor is predicted to drop off without becoming a subscriber.

In this scenario, we’ve leveraged a visitor’s intent of becoming a healthier version of themselves and used that intent to convert them into a subscriber.

Final Thoughts

Identity is core to marketing. Learning to read and recognize context means a more intelligent marketer and a more informed consumer. Inherent in this learning is the ability to make predictions about future behavior, to know the customer more intimately, and to be proactive rather than reactive.