How do you calculate the metric LTV (lifetime value)? How is it relevant in marketing space?

Sr Director of Marketing, JOLYN

The term LTV (lifetime value) has a clear definition for the KPI but I have found it challenging to have one “go to” LTV number to share out to stakeholders when they want to know what is our LTV right now. “Right now” being the key phrase. Maybe I am overthinking this but I have been looking at LTV in context to cohorts to see how each cohort is contributing to our sales and to understand the quality of each cohort over time instead of looking at one number for the state of the business.If I were to abandon that use case, and just wanted a “real time” LTV, I have a few ways of achieving this but wanted to get feedback from the community as to how others are doing this. Also important to note is that the KPI of LTV:CAC is used in these conversations as well.Options:

  • Do a look back a the lifetime, in this scenario, we’ll use 12 months. Average revenue per new customers acquired in the last 12 months minus average COGS. (In this scenario, I can also look at the last 12 months CACs to get the LTV:CAC). This is the equation that I am leaning towards.
  • Look at full database no matter when a customer was acquired and look at the average revenue per customer that has been on the books for 12 months minuse average COGS. Problems with this is that it doesn’t provide good insights over time.
  • Same as option 1 but only measure those acquired in the month that is 12 months ago. In this scenario, I have a smaller cohort but I have allowed for the cohort to repeat purchase every month for 12 months where in option 1, the cohort includes those acquired 12 months ago but also 1 month ago. In this scenario, I could pull LTV:CAC by getting the CAC for the month that they were acquired.

Do any of you need to have this metric on hand and how do you calculate it?

Data Scientist

For methodology sake we still talk about about an LTV value right now for a certain time period (12 months in your case). So the big question is how do we have a relevant values for 12 months of customer lifetime.You obviously put a lot of thought in in the options and are looking to account for the fact that not all customers have had 12 months to demonstrate their behaviour. You’ve already mentioned two options, but to summarize:

  • Ignore this fact - the LTV will appear lower
  • Only look at the customers that purchased at least 12 months ago (since you’d have multiple cohorts here assuming the company exists for 13+ months) allowing for observing changes in time. This results in correct numbers, but those might not describe the behaviour of the newest cohorts
  • Predict the LTV for the remainder of the user’s lifetime based on past behaviour
  • Simpler approach - using ARPU for 30 days and extrapolating based on user retention
  • More complex approach - predicting on user level

I am wondering if there’s space for some education in regards to stakeholders. Sometimes the best we can do is present the implications of the methodology and get more inputs as multiple options are valid depending on business context.

Marketing Analytics Veteran

what I like about option 3 when evaluating LTV is that regardless of which month a new customer purchased, they all have the same 12 months for repurchase. I would raise the considerations of shipping revenue and cost, promotion and marketing costs and possibly return rates depending on how much of an issue it is and whether it is fairly consistent across all marketing tactics. Historically, when I factor particularly promotion and marketing costs into the profit per order and into LTV, I have found the winners and losers vary greatly.

Data Architect

I would frame it as a predictive LTV problem, a bit like how Roman has described it.For example, I would cut LTVs by 1 month, 6 month and 12 month lifetime cuts, and look at various cohorts. Like the screenshot I've attached.Naturally, customers acquired in April 2023 have had 12 months to mature, so they’d have a true 12 month LTV, but to predict the 12 month LTV of customers you acquired in March 2024, you have to 'predict' their LTV based on their activity in 30 days since you've acquired them. So, like how Roman describes it, I would simply calculate the ratio of 1 month LTV to 12 month LTV for the April 2023 cohort, and apply that ratio to the 1 month LTV for the March 2024 customer cohort.See how I've done it below. You could layer the promo costs and other concerns that Ellen has brought up at this point and make it complete.