Cautions about using one industry to create examples for Customer Lifetime Value

Lloyd Melnick has a short and snappy book on customer lifetime value (which he terms LTV). At its best this is a useful insight into the games industry. There is a lot of detail there in a very easy to read form. For example, you will learn about ARPU (Average Revenue per User) and ARPDAU (Average Revenue per Daily Active User). He gives some good advice, e.g., LTV must exceed CPI (cost per install), so there certainly is value in the book.

Where he lost me was that he hopes to give advice across other industries because he says, “LTV transcends all industries and the game industry is a good starting point for everyone to use when learning how to optimize LTV as it is at the cutting edge of leveraging analytics” (Melnick, 2016). I wouldn’t argue that lifetime value is an important concept in many situations. I also believe games companies have superior analytics to many. My problem is the idea that the games industry is a good exemplar. The challenge is that his advice breaks down outside the games industry very quickly. Applying approaches from games to other industries is very dangerous.

Melnick clearly realizes that costs matter. He advises “Incorporating costs and expenses in LTV” (Melnick, 2016). He does this after laying out his calculation approach that largely downplays costs, using revenue rather than profits. His casual approach to calculating lifetime value also downplays discounting. This means his ideas about lifetime value are highly problematic. He might get away with this casualness in the games industry as games often have very high fixed and modest variable costs. The fixed costs are sunk once a game is already produced so revenue may be a pretty close proxy for profit. Similarly not discounting is a terrible idea in lifetime value calculations but may hurt less in gaming as the lifetimes in this industry often make fruit-flies look like Methuselah. With low variable costs and short lives the major challenges he ignores, discounting and costs, matter less to lifetime value calculations. However, many (most) situations where lifetime value is used don’t have these characteristics. Melnick’s advice becomes quite dangerous. This is obvious when you see the example he borrows of Starbucks. The example he gives averages 1) not discounting and not using costs, with 2) using costs but not discounting, and 3) discounting but not using costs. I have absolutely no idea why this is done. It isn’t even simple. Not discounting and ignoring costs is simply a terrible idea for Starbucks given they have reasonably high variable costs and the lifetimes we are talking about are 20 years. Creating some sort of average from a load of bad calculations is just silly.

The takeaway; always use costs and do discount when performing lifetime value calculations, especially if you aim to give advice that transcends industries.

Read: Lloyd Melnick with Wendy Russ Beasley (2016) Understanding the Predictable: How to calculate, understand, and improve customer lifetime value, to build a great company,