One of our customers is navigating the path from a product-centric strategy to a customer-centric strategy. This has a cascade of implications, from how they organize their business, to the compensation of sales teams, to how they think about pricing. Their handbook for this transformation is a wonderful short book by Peter Fader called Customer Centricity: Focus on the Right Customers for Strategic Advantage. This book is a wiry 117 pages and it packs a punch, like Geraint Thomas going up the Alp D'Huez in the Tour de France. You can't afford not to read this book.
Peter Fader defines customer centricity as
"a strategy that aligns a company's development and delivery of its products and services with the current and future needs of a select set of customers in order to maximize their long-term financial value to the firm."
There are two key phrases here: 'a select set of customers' and 'maximize their long-term potential value.' These two ideas are closely connected. The select set of customers is the set of customers that will maximize the long-term financial value.
In one of the most interesting parts of the book, Fader looks at a common mistake that people make when calculating customer lifetime value. This caught my attention as customer lifetime value (LTV or CLV) is something we use at Ibbaka in our segmentation and customer targeting work, and we are often asked to predict the impact of price changes on this metric (we do this using Monte Carlo modelling).
His basic point is that when you calculate an average lifetime value for all of your customers, without segmenting first, you get the wrong number for average lifetime value and this can lead to poor investment decisions. Here is the example he gives.
If you simply take the average churn across all customer you would calculate (0.06 x 0.70) + (0.35 x 0.20) + (0.65 x 0.10) = 0.177 (or 17.7%). This give s an expected customer lifetime of 1/0.177 which is 5.6 years. This number is wrong.
What you should do is calculate the expected customer lifetime for each segment separately and then take the average of this. Doing this we get (16.7 x 0.7) + (2.9 x 0.20) + (1.5 x 0.10) = 12.4 years. So the actual average customer lifetime is 12.4 years and not 5.6 years, a difference of more than 2X.
Let's watch this play out over time. Here we have graphed the number of customer over five cycles (the cycle will be whatever was used to measure attrition rate, which in this case is years).
Now I find this distribution of low risk to high risk customers rather optimistic. Let's try a more pessimistic assumption.
Under this distribution, the customer drop off is much steeper.
Let's layer in price and look at some different scenarios. In the first case we keep the original distribution of customers and look at two different pricing strategies. In strategy 1 the company has decided to give the low risk customers a low price to reward them for their loyalty, and to charge the high risk customers a premium to capture value from them before they leave.
In strategy 2 the company does the opposite. It prices an offer for the low risk customers higher as it is able to deliver these customers more value. The high risk customers select the lower priced offer.
Note the implicit assumption that the offers are not the same. Segmenting on probable attrition rate (churn) and pricing strategy leads to different product strategies. And the product strategy depends on the differentiation value for each set of customers, more on this below.
Let's check these two strategies under different distributions of high, medium and low risk customers. As we did above, let's make the more pessimistic (realistic?) assumption that only 10% of the customers are low risk, 20% are medium risk and 70% are high risk. Here is the cohort revenue contribution under strategy 1 (where the high risk customers get the higher price) and under strategy 2 (where the low risk customers get the higher price).
Summarizing the outcomes for these two distributions and pricing strategies we get the following table. The Y axis shows the two different distributions on customers (the top row is for the distribution where there are more customers at low risk of churning, the bottom row is where there are relatively more customers at high risk of churning). The X axis represents the two pricing strategies (in the left column the pricing strategy is to charge customers at high risk of churning a premium; in the right column the customers with a low risk of churning are charged a premium). The numbers in the cells show the revenues for the cohort over six revenue cycles. These numbers are based on a the specific distributions used and are actually quite sensitive to the distribution assumptions. In the real world one would use a Monte Carlo model to explore different outcomes.
These results may seem counter intuitive, but in fact this represents two easily observed strategies.
1. Get lots of new customers, milk them, accept churn
2. Focus on the low risk customers, nurture them, but make sure you are well paid
The focus here is on the value of the customers to the vendor. At Ibbaka we look at this from a different angle. The set of customers that will maximize the long-term financial value is the customers for whom you create the most differentiated value. This is why we begin by segmenting the market by value drivers (emotional and economic) and buying process. We then help our customers target the segments where they create the most differentiated value. These are also the customers that will be low risk. There is a set of positive feedback loops between the customers for whom you create the most differentiated value, the companies that are at the lowest risk of churning and the customers that you can charge the highest price. It is this set of customers that you want to target. Value-based market segmentation and value-based pricing fit well with customer centric business models.