The concept of Lifetime Customer Value (commonly referred to with the acronyms LTV or LCV) is a framework for thinking about freemiumproduct unit economics. A "LTV" metric is an estimate of the total revenue that the average user in some group will contribute to a product over the entire course of their use of it. LTV can be thought of as an expected value of a user.
My personal approach towards LTV utilizes the concept as a marketing metric — it is a means of breaking out the user base into dimensions that are functional from a marketing standpoint for the purpose of setting bid targets. For instance, you might have a LTV metric for users acquired on Facebook based in the US on iPhone (for a mobile app) or for users acquired via email referral from an existing user based in Germany (for an e-commerce website).
LTV metrics are generally calculated by measuing the cumulative daily revenue contribution of cohorts over time and then using some set of statistical tools to project those contributions forward. Many product LTV curves take a logarithmic shape, or seem to approach some asymptote, over time as users in cohorts churn out and per-user revenue contributions (which are measured over time on the basis of the entire cohort, not just those alive at a given point of time) decrease per day. An example of such an LTV curve, with cumulative per-user revenue contribution on the Y axis and days since the cohort joined the product on the X axis, might look like this:
There are myriad ways to dimensionalize LTV, but if it is being used as a marketing metric, it makes sense to segment users on the basis of marketing channel parameters that are configurable / targetable by the marketing team, otherwise the values of those LTV metrics won't be useful for the purposes of marketing to new users. For instance, I could theoretically break out user LTVs by the marketing creative they saw when they clicked on an ad for the product, but if those creatives are being used across multiple campaigns across multiple geographies and languages, then this creative-based LTV isn't useful to me in optimizing campaigns. I'd need LTV broken out at the most granular targeting set for a campaign in order to use it to target that campaign.
Once LTV is broken out in a way that is useful for the marketing team, then it's fairly easy to utilize that number to optimize campaigns. For instance, using the Facebook users in the US on iPhones example from above, if the estimated LTV for users in that group is $5, then any users acquired for less than $5 are profitable over their lifetimes in the product for that campaign.
But this reveals the struggle in using LTV productively: marketing costs in user acquisition are realized upfront and LTV is paid out over some course of time. So for the most part, the wholly conceptual "LTV” metric is broken out by companies into timeperiods that make sense for them to monetize against given their financial wherewithall. Some companies might buy traffic against a "90-Day LTV,” meaning that they have calculated the expected revenue contribution of users in 90 days and intend to have their marketing costs for those users repaid via their revenue contributions in their first 90 days in the product (with all revenue contributions after that being marketing profit for those campaigns). Some companies use the "Day-X LTV” formulation for describing this concept and some use "Day-X ARPU (Average Revenue Per User)."
The appropriate timeline to use for setting bid targets on the basis of expected revenue contribution is a function of how a number of different characteristics of the company and the product. If the company is sitting on a massive balance sheet and doesn't need to return cash from advertising campaigns very quickly in order to fund operations, it might buy traffic against a longer LTV timeline (eg. 180 or even 360 days). If a company has very little data on how the LTV curve materializes over time and wants to be cautious around marketing profitability, it might use a short LTV timeline at first (eg. 30 days) and extend that timeline out as it gathers more data and sees its LTV curves evolve. And if a company is operating in a very competitive category in which marketshare can evaporate quickly, it might likewise operate on a short LTV timeline with the assumption that users could defect to other products, meaning that LTV curves change as a result of exogenous market factors.
Some companies put a lot of effort in building very complex, highly-parameterized models to calculate LTV, but in my experience, the most effective and robust models are simple and elegant. Since most freemium product monetization is driven by outlier behavior, the more a model is parameterized, the more likely it is to overfit the on the basis of high-engagement users in specific cohorts in a way that doesn't generalize to future cohorts. A simple curve fit exercise, in my experience, tends to yield the best insights into how user-group LTV evolves over time.
And conversely, I've also generally seen companies under-invest into the tools and processes that actually utilize LTV for commercial impact. For instance, if an LTV model is proven to be reliable and trustworthy, it provides incredible guidance on daily cash flows: companies tend to not use their LTV models to optimize marketing spend on the basis of revenue recoup and cash management. If I know what my cohort LTVs are and I know how many users I am acquiring every day, I effectively know how much money I'll make each day on a forward basis; if that's the case, I can very systematically plan marketing spend while maintaining some minimum cash balance without having to resort to overly simplified cash management techniques such as only spending a fixed percentage of monthly revenue on marketing.
For most freemium products, LTV is really the cornerstore of their commercial operations: it dictates the extent to which the product can grow through paid user acquisition. It makes sense for companies operating freemium products to invest heavily into building business processes and infrastructure around their LTV analytics and to circulate that insight throughout the company, even though I firmly believe that LTV should only be actioned against by the marketing team. But a product team should see the LTV metric as a valuable indicator of where they need to focus their attention: if the LTV / CAC (cost of acquisition) dynamic is unfavorable, it means that the product can't be marketing efficiently because of some combination of: 1) it is not approachable enough to compel people to click on ads for it, 2) it doesn't retain well enough to drive long-term monetization, 3) it doesn't monetize well enough to deliver meaningful per-user revenue impact. LTV isn't something that a product team can action against, but the LTV / CAC dynamic can provide helpful guidance to the product team into digging into one or many of the above product issues.
