Customer Segmentation Strategy: The Importance of The Right Message

As marketers, you know that personalization and relevancy are crucial to reaching today’s audience. There’s a lot of emphasis on 1-to-1 personalization, and it’s a concept everyone seems to understand. But what does the concept really mean when it’s time to deliver that 1-to-1 messaging?

Historically, it’s been going into an email service provider (ESP) and manually building segments that make some intuitive sense: High spenders who live in LA; or females who bought in the sports category in the last 30 days. Then target these segments with some specific messaging: “Damn it feels good to be a VIP in LA!” Then customize this message in any number of ways with variables such as {username} and {user location} and {sale date} and {rewards balance} and {product interest}. And perhaps create a series of emails to give these users multiple touches.

This is a fine way to use ESPs, as they provide deep segmentation and flowchart capabilities to enable an intense amount of manual personalization. But is there a better way to do 1:1 personalization that is quantifiable?

For example: Within your segment of LA VIPs, were some of them churned? For the females who bought in the sports category, could they also be interested in high fashion? Does everyone need information regarding their rewards balance, or their own location, and other things you’re displaying to them?

Confused-Baby-Meme-Blank-20

You could segment further and customize more messages, but this means coming up with a potentially endless amount of micro-segments yourself. This is just about as unscalable as it sounds, not to mention a massive time sink for something that is typically prone to errors.

Just how error-prone is manual segmentation, and why? Oftentimes, manual segmentation turns into a neverending loop of looking for different segments, instead of really evaluating what stage the customer is in — and what kind of message best addresses that stage. We looked at a few promotional email blasts, and evaluated what would have happened if our Artificial Intelligence engine were given the opportunity to make the decisions instead.

For a campaign over a two-day period, 53,837 users were blasted a promotional message that ended with a .04% conversion rate. That’s bad, not even factoring in these blasts’ drastic effect ondeliverability.

Our Artificial Intelligence engine determined that out of the 53,937 users, 47,238 were in a state ofNeeding Nurturing and incorrectly received the promotion, while 6,699 were Ready to Buy and correctly received the promotion. Customers who Need Nurturing tend to be less receptive to pushy content, and need more thoughtful engagement. We then looked at the performance of these two groups:

Screen Shot 2016-03-08 at 7.00.58 PM

Users who Needed Nurturing converted 9.5x worse than Ready to Buy on a promotional email.

We then looked at 3,229 sends that went to the Needs Nurturing customers that were less promotional and focused more on engaging content, while leveraging our AI frequency and timing predictions.

Screen Shot 2016-03-08 at 7.20.08 PM

The right message at the right time converted 48x better! And even better than the promotional emails to promotion-ready customers. Is this where 1:1 personalization should be focused? Let’s drill down more.

We went back and looked at the original 47,238 again, and wanted to see which users our AI predictively scored as Churned or At Risk of Churning at the time, and how they performed relative to Ready to Buy users.

Screen Shot 2016-03-08 at 7.09.42 PM

Just a little bit better than the Needs Nurturing group as a whole, but not by much. And surprisingly, promotional content performed equally poorly across both At Risk and Churned. But how did more appropriate “winback” type messaging work with some of these At Risk andChurned users, again with AI-driven frequency and timing predictions?

Screen Shot 2016-03-08 at 7.28.09 PM

Screen Shot 2016-03-08 at 7.29.52 PM

Churned users converted 24x better! But more importantly, the time-sensitive At Risk customers converted 40x better. And At Risk users are notoriously difficult to win back once they completely churn. The 17% drop off in conversion rate between the two groups shown above is proof of that discrimination.

So how should you take advantage of this? Run your customer base through a platform that doestrue data science, and use true user predictive scoring to target your 1:1 personalization based on where your customers are in their lifecycle. This can be easier than you think with the right tools. Leveraging Artificial Intelligence to manage your users doesn’t have to be as complicated and daunting as it seems, and you and your business will quickly reap the benefits.

The Future of Marketing Lies within AI | Retention Science