One of the most important metrics for any subscription business is customer lifetime value (LTV). This is the value of the revenue one can expect from a customer over the full life span. This metric was popularized by mobile phone companies, who needed to know how much a subscriber was worth so that they could know how much to invest in winning and keeping a customer. Cable companies also ask this question, and they are especially interested in knowing what combination of services (content mostly) will optimize lifetime value.

Modern SaaS (Software as a Service) companies obsess about this metric. It is one of the key ways they are measured by their investors. In fact, there have even been recent acquisitions in which the total customer lifetime value was a factor in the valuation!

For most of us, it was David Skok’s now classic articles that taught us the basics.

SaaS Metrics 2.0 – Detailed Definitions

Our smartest SaaS customers ask us to design pricing to optimize Net LTV. How do we use these calculations in our pricing work? To do this we have to do a number of things:

Segment the customer base on LTV
Model the impact of price on ARPA
Model the impact of price on Churn Rate
Work out the interactions between ARPA and Churn Rate
Consider the impact of price on Customer Acquisition Costs

Segment the customer base on LTV

Last week we wrote about SePVA or Price Dispersions in Critical Metrics for Pricing Execution - Insights from Dick Braun. If you are in a subscription business, it is critical to segment your customers by LTV. Not all segments are equal and not all customers make the same contribution to LTV.

Start with your estimate of LTV for each customer on the X-axis. It is worth doing this for both Gross LTV and Net LTV. Then experiment with different values for the Y-axis. You are searching for a Y-axis that organizes the LTV distribution in a meaningful way. This could be that the customers distribute themselves onto a nice line of some type with a high r-squared. Or it could be that the customers cluster into well defined groups. It is worth looking for both outcomes.

Here are some things to try for the Y-axis:

Churn - It may seem odd to graph churn as churn is a factor in LTV. Some correlation is to be expected. What this does is let you look at the interactions between ARPA and Churn (of course you can also graph ARPA to Churm)
Gross Margin - This is another of those funny ones that can provide surprising insights. Especially when you get clusters rather than correlations.
Engagement - Are your most engaged customers the ones with the highest LTV? If not, why not? What is the relationship between Engagement and Value?
Customer Acquisition Costs (CAC) - This is one of the other critical metrics. Investors will often ask about the ratio of LTV to CAC and will be looking for a ratio greater than 3. Are the customers with the highest CAC the ones with the greatest LTV? If not, is there something wrong with your marketing allocation?
Value to Customer (V2C) - Value to Customer is one of the most important SaaS metrics but it receives little discussion. It is an estimate of how much value the solution is providing to the customer. If LTV > V2C then you do not have a viable business model. Does LTV go up with V2C and is this linear, or some other function (in the real world it is almost never linear).

If there are distinct segments for LTV, and there usually are, you need to think through how to treat each segment. A simple framework for doing this is presented below.

Model the impact of price on ARPA

Once you have segmented your customers by LTV (and V2C if you can) you need to understand the impact of price on ARPA. At the simplest level, price impacts ARPA in three ways.

Price is one input into revenues, as revenues are basically just price x units.
Consider price volume tradeoffs for each segment. Will an increase in price cause the number of units to drop by an amount greater than the aggregate LTV for the segment? Will a price cut (or discounting or a special ‘campaign’ price) generate enough new units to make up for the reduction in price? These behaviors are generally very different for each segment, which is why you need to complete the segmentation first.

Model the impact of price on Churn Rate

Price also impacts the other key input into LTV: the Churn rate. If prices are too high, then churn will increase. This is especially true in for cases where LTV > V2C or even for LTV ~ V2C. It is often useful to construct hypotheses on the impact of price on churn and then to test them on historical data. When we have done this we have generally found only a week correlation between price and churn.

But before you change the renewal price, look at all of the other factors that impact renewals (renewal is just the inverse of churn).

Are customers engaged?
Are customers getting value?
How has the customer’s business changed?
How has the competitive landscape changed?

Yes, one can raise prices so high that churn increases. If this is the customers in the lower left quadrant above, who are provind low LTC and receiving relatively little value, then churn could be a good thing for both parties. In most cases, price is not what is driving churn.

Work out the interactions between ARPA and Churn Rate

ARPA and Churn are not independent of each other. Remember that churn can be greater than 1. This means that you are able to grow your business with your existing accounts. A price increase can

Increase ARPA if the additional revenue from the price increase is greater than the loss in revenue from lower unit sales;
Improve Revenue Churn if the additional revenue for current customers from the price increase is greater than the loss in revenue from increased account churn;
or
Decrease ARPA if the additional revenue from the price increase is smaller than the loss in revenue from lower unit sales;
Increase Revenue Churn if the additional revenue for current customers from the price increase is smaller than the loss in revenue from increased account churn.

Build a model that factors in the effect of price changes on both ARPA and Revenue Churn. Look carefully at the feedback loops and interactions between the impact on ARPA and the impact on Revenue Churn. Test if the model works the same way in each segment.

Consider the impact of price on Customer Acquisition Costs

Price can also impact Customer Acquisition Costs (CAC). The higher the price the higher CAC are likely to be. In our experience, this continues on into customer success investment. High value, high LTV solutions most often come with high Customer Acquisition Costs and require an ongoing investment in customer success.

On the other hand, low Customer Acquisition Costs and low subscription prices are often related to higher churn rates. As price is based on value to customer (if you want to have a sustainable business) one needs to triangulate between Value to Customer (V2C), Customer Lifetime Value (LTV) and Customer Acquisition Costs (CAC) in designing subscription pricing.

A more complicated story…

It is a challenge to manage a subscription business and make sure that revenues grow over time. It is much more complicated when a company has two or more offers that can (i) compete for customers but (ii) feed each other customers. In this case one needs to have a clear (and shared) understanding of the role of each of the offers. (For some context on this see Pricing Your Solution Portfolio Part 2: Setting Goals). Once you have alignment on goals, build a predictive model for how different pricing models will impact V2C, LTV and CAC and, here is the fun part, model the interactions.

Take CAC as an example. If one service feeds another it is effectively reducing CAC costs for the second program. It is also providing pipeline. A change to pricing for one service will cascade into the CAC for the second solution.

There are also interactions between LTV. A change in prices for one service will normally cause some redistribution of subscribers between the two solutions. In other words, it will increase churn for one of the solutions, which of course impacts the LTV. Depending on the role and strategic goals for each solution, one can build an optimization model. This model should be predictive. The initial predictions will probably not be all that accurate, but as you gather data and run experiments accuracy will improve over time.

These kinds of interactions also exist in multi sided markets, and building models for this sort of pricing interaction is part of our ongoing research.

Here we have focussed on subscription models, but these concepts are also relevant to other pricing models. Any business where you are building a long-term relationship will benefit from framing pricing strategy in terms of customer lifetime value and the other key metrics for unit economics.

Pricing for lifetime customer value