Search the website

The Beauty of Customer Segmentation, Part 3/3: Let Your Customer Segmentation Model Do the Talking

There's no secret sauce here. Not all strategies will work the same for everyone all the time. But you can still identify the optimal messaging strategy for your customers. Just listen to what your customer model says

Posted in

Welcome to the final part of this mini-series! In the first two parts, we detailed why combining online, and offline data can take your marketing initiatives to the next level and defined both role-based and cluster-based segmentation:

The Beauty of Customer Segmentation – Part 1/3
The Beauty of Customer Segmentation – Part 2/3

Today, we’ll turn our focus to what your customer segmentation model is telling you – as it’s a powerful source when it comes to determining what messaging and strategy you should use when marketing to your customers.

There’s no secret sauce here; we’ll tell you that right off the bat. Strategies that work for a particular brand and customer may not work for another, and techniques that worked for you in the past may not work as optimally in the future. And so on.

Having said that – how can you possibly identify the optimal messaging strategy to deliver to your customers at scale?

Well, that’s where the segmentation model comes in. Not only is it ideal for sending highly-personalized and relevant messages, but also, it can be used more strategically as it can expose valuable marketing strategies to increase retention and customer lifetime value (CLTV).

Let’s take a more in-depth look at it. You’ll see it’s as easy as 1,2,3.

1. Divide Customers into Lifecycle Stages

At Optimove, we first divide customers into lifecycle stages and then add additional segmentation layers – clustering customers under each segmentation layer based on their behavior.

To determine what layers to use when creating the segmentation model, a dedicated data scientist reviews a brand’s customer data, to identify which behavioral patterns expose valuable insights. These insights can help marketers understand who their most valuable customers are, how often they purchase, what behaviors lead to the highest predicted future value, and what behaviors they exhibit before churning. 

For example, let’s take a campaign in which you want to offer some sort of promotion to your customers. You can take one approach to test two different treatments, such as “buy one get one free” against “20% off your entire purchase”.

Both strategies offer a relatively high promotion to customers who may not need such a high incentive to purchase. Alternatively, you can examine your customers’ average future value based on the various factors or the layers in the customer segmentation model to locate those with the highest future value. You can then start exploring additional customer attributes that make up the cluster with the highest future value.

Once you’ve identified the cluster’s customer attributes with the highest future value, you can start creating a strategy that encourages similar behaviors in segments with a lower average future value.

For example, divide your customers into clusters based on their discount affinity, as seen below. In this case, customers who received no discount or very low discounts have a higher future value than those receiving heavy discounts. Therefore, to increase the customer’s future value, you can encourage customers who’ve purchased with heavy discounts to do so with lower discounts or full price.

2. Identify Customer Activities

Marketers should identify which activities new customers with the highest predicted future value demonstrate in order to encourage similar behaviors in other new customers.

You can see that customers with few high-value purchases have the highest future value in the table below. Consequentially, you wouldn’t necessarily encourage customers to place many orders, but rather fewer orders of higher value. A good strategy would be to trigger a website pop-up before the customer completes the order, offering promotions on items that complement those in their shopping cart to increase the order value. (A classic upsell mechanism!)

Customers who have few high orders and are already in the cluster with the highest predicted future value probably don’t require heavy incentives. Offering them one would likely result in lower ROI to the marketing campaign.

To avoid profit margin cannibalization, offer different levels of promotion to each of the clusters. Those with higher future values should be offered lower incentives, while those with lower future values should be offered heavier incentives.

3. Scale Your Offerings

When you’re sending out dozens or hundreds of different campaigns, it’s difficult to maintain and monitor various offering levels for different clusters.

To make matters simpler, Optimove customers use the Self-Optimizing Campaign to set different levels of promotions to A/B/n treatments. The Self-Optimizing campaign detects, and autonomously delivers the campaign that will drive the highest CLTV to each customer. 

Using Self-Optimizing Campaigns not only reduces the tedious task of manually selecting which promotion to send to each cluster but can also detect preference shifts and autonomously adjust the treatments. 

You can read more about the importance of being able to adapt marketing strategies to shifting customer expectations here.

Conclusion

Coming up with a retention marketing strategy isn’t a simple task and identifying the winning message is nearly impossible. However, if your strategy is data and insight driven, you’re more likely to land a winning approach. 

So next time, when you’re planning out your marketing strategy, take a look at what your customer model is telling you, it may just spell out the exact marketing strategy you need.  

 

Published on
Posted in

Gabriella Laster

Gabriella Laster is Product Marketing Manager at Optimove, responsible for prospect marketing and product messaging. Gabriella holds a B.A. in International Relations and English Literature and an Executive MBA from Hebrew University.