By: Ram on Jun 19, 2022
Customer acquisition is one of the costly activities and most important for them to grow. “Most sites spend between Rs 800-1,500 to acquire a customer”. And the expectation is that high % of these customers will repeat their first purchases.
CLTV can help in identifying the prospects (potential customers) who make high value for an ecommerce and then find the customers who are more of high value customers. This helps the organization in building future value pipeline from the current acquisition initiatives.
In this blog, we describe our approach on how ecommerce organization can build the Future Value pipeline using Customer Acquisition. It has 3 high level stages and each of these stages are driven by Machine Learning or Data Science techniques.
As described in the previous blog, we can consider a foundational metric and period for CLTV and calculate the CLTV for each of the customers.
where CLTVn is CLTV measured for each customer (1,2, 3…n)
There are two important applications of CLTV Measure or estimation for the new customers
We will consider 6 months total value (in term of foundation metrics) and then apply a multiplier factor for the next 18 months.
We have considered 3 segments – High, Medium and Low Segments. For the historical cohorts, segmented the customers into these 3 segments and calculated for next 18 months CLTV.
Let us consider this with an example where, for High (Top 30%) and Medium (next 50%) CLTV segments, the value grown by around 30% in the next 18 months (post first 6 months) and for the Low (bottom 20%) value segment the value is further decreased by 10%.
Final calculation for CLTV Measurement at the customer value.
We have CLTV value at the customer level and now we need to build “Value Driver Model”.
We will develop a Machine Learning (ML) Model that will help us identify the drivers of CLTV. There can be multiple ways to develop the ML Model Framework.
We will discuss the second approach – ML Classification Model. The CLTV estimated in the CLTV Measurement stage is used to split the customers into 3 groups – Top 30% as High, Next 50% as Medium and the last 20% as Low Value segment.
Now, we have defined the outcome or target variables. And the next step to hypothesise and consider all the features or characteristics that can help in differentiating High Value customers from other classes.
Some of the key features that can be considered are above. We need to prepare the data and above features to start the Machine Learning (ML) model development journey. The technical process of the model development will be discussed in subsequent blogs.
The Driver Model that helps in identifying the key characteristics of the high value customers. As we expect there will be multiple behaviours that drives the high value to the ecommerce company. And we need to identify the profiles of these sub-segments. The profiles of these subsegments will be relevant in multiple ways.
The prospects who are look a like of ‘High Value” customers are likely to bring high value over the lifetime with the ecommerce organization. The sample of these sub-segments can then be submitted to platforms to acquire look alike customers.
Based on the offer relevance characteristics, the specific and relevant offer construct can be created for each of these sub-segments. And the same will be communicated when targeted on the relevant platform.
It will be evident from the subsegment characteristics on their associations toward a category or brand. The same can be leveraged for the targeting.