When evaluating which customers are most attractive to your business, you can use different methods to segment customers and leads. In the long run, one of the most important things to look at is CLV – also known as Customer Lifetime Value. The name speaks for itself; it is a measure of how much you can expect to earn from a customer over the entire duration of the customer engagement. Particularly in connection with acquiring new customers, it can be useful to have an idea of different customers’ CLV, as it needs to be weighed against the cost of acquiring customers, also known as Customer Acquisition Cost (CAC). Some variations of CLV even factor CAC into the equation.
A calculation of a customer’s CLV could, for example, be the following: [Number of purchases per year] x [Avg. purchase value in currency] x [Avg. profit margin] x 1/[churn rate* per year] = Customer Lifetime Value
*churn rate is the proportion of customers who leave you within a given period
In the past, it was only possible to estimate CLV for existing customers and then divide them into segments to target marketing efforts at customers with the highest average CLV. But with the use of Machine Learning and CDP (Customer Data Platform), the possibilities for estimating and predicting future CLV have advanced significantly in recent years.
Developing CLV models based on regular and comprehensive analyses of your existing portfolio is typically extremely time-consuming and often relies on more or less subjective segmentation criteria. Instead, you can develop models that predict future CLV at an individual customer level based on available data. This provides a more detailed and streamlined method, which can easily be updated as the market changes.
For many years, both common practice and academic research in CLV prediction have focused on looking up statistical distributions, which, to simplify a bit, were based on the traditional idea of the Pareto principle: that 20% of customers account for 80% of sales. It was relatively simple to collect data for such models, as they only needed data about the customer’s purchases based on the RFM criteria (Recency, Frequency, Monetary value) – but in return, it required significant purchase history from the customers. These models also relied on assumptions that customers purchase products in a fixed pattern and at a constant frequency, such as supermarket customers.
However, the evolution of e-commerce with highly variable purchase patterns (and in some cases fewer but more lucrative customers) has required other methods to assess which customers have the greatest potential. At the same time, the internet enables the collection of far more detailed data about customers, rather than just basing models on transactions. This is where Machine Learning comes in. Machine Learning algorithms, using various mathematical methods, find patterns in data that can explain specific variables based on various inputs (e.g., transactions and demographic data). In recent years, predictive CLV models based on Machine Learning have begun to outperform both simple segmentation models and the aforementioned statistical distributions by a significant margin.
Existing customers with multiple purchases: For these customers, it is often possible to calculate a fairly accurate CLV, as they have enough purchase history to form a clear pattern. In this case, it is often sufficient to know the average churn rate, purchase frequency, and order size for the customer segment.
New customers: It is often interesting to segment customers the second they make their first purchase. However, it is difficult to assess a customer based on a single purchase, as you cannot gauge how loyal they will be or which other products they might choose in the future. However, demographic information based on a customer’s address can often provide significant insight, and ML models have shown that knowing the customer’s first purchase often makes it much easier to predict their future behavior, as there are similar customers to compare them to.
Leads: Leads are potential customers who, for example, have visited your website and subscribed to a newsletter or similar. Demographic information (based on addresses) or behavior on the website can help segment these customers based on their relative potential. Of course, segmentation will not always be perfect, but you will have a good overview of the likelihood of turning them into valuable customers and can roughly sort the attractive ones from the less attractive ones.
Traditionally, companies have often used different rules to segment by CLV, but this is time-consuming and typically requires a lot of prior data analysis to identify attractive customers or leads. With Machine Learning, however, the data tells the story and comes up with the most effective rules. The algorithms simply need to be fed with enough data from existing customers to identify the patterns – often a few hundred data points are enough.
They also have the advantage that, once the setup is running, it requires minimal maintenance – the models learn from your data. Machine Learning isn’t a crystal ball that can predict the future with 100% accuracy (that would require a lot of data), but when prioritizing customers and leads by potential, even grouping them into categories, such as red, yellow, and green (with green being the most attractive), can significantly increase the effectiveness of campaigns. Often, a small number of customers generate a large portion of revenue, and if you can increase the likelihood of creating such “super customers” by 30% in a campaign targeting several hundred recipients, the costs will quickly pay for themselves many times over.
Would you like to hear more about your company’s opportunities with AI and CLV strategies? Reach out to Jeppe Berggreen, CEO
jb@ambition.dk
+45 53 54 55 66
To get the most out of Black Friday, you should not only focus on spending money on ads but also invest time and energy in optimizing your campaigns across channels and seeing what’s hidden in your database.
Leverage user data in GA4 to achieve more precise conversion data here as well.
Danske Licens Spil is powering a multitude of exciting digital experiences this summer including events such as the European Football Championships.
