Lifetime value is obviously important for running a successful business, so how can it be calculated?

I’ve had to calculate LTV many times over my 11 years as a data scientist, so I know there’s more than one way depending on the quality you need (and that depends on how much money you want to make, better calculations tend to make more money).

Or quarterly,

### LTV = average revenue per quarter / churn % per quarter

Or monthly: LTV = average revenue per month / churn % per month

#### Recommendations

I recommend starting with yearly, because it compresses information more than quarterly or monthly (thus lowering the error caused by variance, eg- if there’s a sudden spate of newly acquired customers).

The quality and reliability of the estimate can certainly be improved, though. The limitation of this approach is the risk of using a stale average revenue figure. For example, if average revenue has been increasing (or decreasing) over time, and you want the figure to apply to new customers to drive acquisition, it would be advantageous to use more recent customers’ average revenue per month (but not their churn figures).

The reason not to use churn figures this way is that usually new customers are more fragile than long-time customers (before they “burn in”) and so using new customer churn is likely to overestimate the churn rate (and that would cause one to underestimate LTV and under-invest in acquisition).

The point is that it’s valuable to think about how average revenue and churn have changed over the company’s lifetime.

At this point, the limitation is that the calculation produces a single number, but realistically the actual realized values will be above or below the figure. So how much? Figuring out if it’s a material amount requires taking random samples of customers (with replacement), and calculating LTV as above to produce multiple figures. Then one can take the min and max to learn about the actual range likely to be experienced. This is important for protecting against downside risks caused by over-investment in acquisition, but also for protecting against under-investing in acquisition and causing disappointing revenue growth.

So now we have a trend-adjusted range for LTV. In my experience this is better than most companies can do and can be extremely valuable, but there’s often a lot more ROI to be gained by continuing to improve the LTV calculation’s quality. This is where simple calculations run out of road.

For example, by using advanced econometric and statistical methods, I can create a model that uses fewer and more reliable assumptions. That means fewer ways for the calculation to fail and a higher degree of accuracy. With a higher degree of accuracy, it’s possible to do all kinds of valuable things like: calibrate CAC more finely, invest more in advertising and generate more revenue. Many decisions obviously depend on LTV and that makes it all the more valuable to have an accurate estimate for business. If this interests you please get in touch!

@statwonk