Lifetime Value (LTV) as a Key Metric of Success

• LTV is a key metric of success for any business model geared towards growth
• Its signficance is most pronounced for subscription-based business models
• 41% of new customers begin as regular subscribers
• LTV here is vital as it captures future potential earnings over a substantially long horizon

As a heavily subscriber based business model, LTV is arguably a more vital metric than gross margin. Rapid gains in LTV have been made since 2015 but what appears to be a plateau in the growth of total customer LTV suggests an active effort to increase total LTV is warranted. Two options are available to increase total LTV:

1. Increase the rate of new customer acquisitions
2. Increase the average LTV per customers (i.e. attract higher quality customers)

As customer acquisition costs are rising quickly, increasing the quality of new customer acquisitions is paramount in both increasing total LTV and reducing costs associated with more aggresive broad based customer acquisitions campaigns.

Much of the growth in total LTV is due to an increase in new customers and not to an increase in average customer LTV, as the average customer LTV has changed little over 4 years. Attracting better customers can help boost the average LTV (and overall business health). But what makes a customer “better”?

Customers with the highest predicted LTV tend to spend more on their first purchase and in total during their first 90 days. They also tend to:

1. hail from the Pacific region

2. prefer pizza alone over pasta

3. prefer headphones to other electronic accessories

4. to be subscribers, and

5. to be koala in type

To drive growth in average customer LTV, focusing efforts on targeted customer acquisition along such profiles is key. Specifically, increasing the share of first-timers that are subscribers appears to be a tailored way to maximize LTV.

What sources and mediums have been responsible for channeling the highest quality customers? This is helpful in identifying ways to acquire more of these customers. Google is a clear winner, with Facebook a close second. In fact, these cost-per-click sources represent the medium with the greatst aggregate LTV.

Adjusting for the effects of other variables, the following are the greatest predictors of LTV:

• Customer Type: Pigeons are associated with the highest LTV relative to other customer types
• First Order Type: Subscribers are predicted to produce $51 more in lifetime value than one-time purchasers
• Product Preference: Those who prefer headphones are predicted to have higher LTVs than those who prefer alternatives
• Food Preference: Those who prefer pizza alone are predicted to have higher LTVs than those who prefer alternatives
• 90 Day total revenue: each additional dollar spent in the first 90 days is associated with ~$3.5 additional in LTV
• Region: The West is a predictor of higher LTV relative to other regions

These predictors allow for a more tailored approach to customer acquisition campaigns

• More capital can be invested in marketing in the West
• Promotions can be rolled out that push first-time buyers to subscription services over one-time purchases
• Discounts to drive first 90 day revenue streams

“We already use 90 day revenue as customer value… Why should I use LTV instead?”

1. For the same reasons that a 90 day revenue window captures better the hazard of customer attrition than raw sales do, LTV captures a longer profit horizon tuned to such tenured attrition. This allows businesses to calculate longer-term ROIs, which are vital to the subscription service model.

2. Because of its narrow temporal range, inferences drawn about future customer behavior and value from a 90 day revenue window are are unreliable at best. LTV, because it captures a far longer time horizon, is a more sound metric to organize strategy around.

3. Variation in behavior b/w customers is more discriminating as time passes. 90 days is too short a time period to reveal stable customer segments.

“Retina fit a linear regression model to predict LTV and it produced an R2 value of 0.68… Is this good? Why?”

R-squared tells us how much of what we want to explain are we capable of explaining with our model. But reliance on it for model assessment should be undertaken with caution.

LTV as an example: 1) the relationship between LTV and its predictors is nonlinear (as predicted LTV should not dip below 0, a linear model will force this), in which case R-squared will be misleaing and 2) without context, the value of R-squared can be difficult to anchor to any single conclusion -good or bad, linear or not. That context includes the decision-making situation, one’s objectives and goals, and the subject matter context.