Understanding and Estimating Lifetime Value (LTV) in Business: Concepts, Methods, and Applications

1 Related Concepts

LTV (Lifetime Value) represents the total value generated by a user over their entire lifecycle. Historically mentioned since the 1980s, its definition varies: some treat it as net profit after removing costs, others as the discounted net present value of all future profits. In practice, Liulishuo combines LTV with CAC (Customer Acquisition Cost) to compute ROI, while LTV itself is defined as the total revenue contributed by a single user, without cost deduction.

2 Estimation Logic

LTV must be estimated using methods suited to the specific business context. For Liulishuo, user value is linked to the sequence "registration → first purchase → subsequent repurchases → churn". The estimation proceeds in two steps:

Step 1: Estimate the long‑term first‑purchase conversion rate by applying linear regression to historical "registration → first purchase" data, obtaining a regression coefficient and using the current first‑purchase rate as a baseline.

Step 2: Estimate the LTV of first‑purchase users. SKUs are categorized, and the purchase‑and‑churn pathways are displayed as a multi‑branch tree. Historical data provide the Average Order Value (AOV) for each node, while the node probability equals the estimated repurchase rate (recent repurchase rate × repurchase regression coefficient).

The figure below illustrates the LTV estimation logic for a single‑type first‑purchase SKU with two repurchases.

In practice, multiple SKU types and additional repurchase cycles are handled using the same logic.

3 Model Evaluation

LTV prediction spans a long horizon, and the longer the period, the closer the forecast aligns with reality. As shown in the following chart, predicted LTV (predicted_ltv) and actual realized per‑user value (real_value) from 2018 and earlier exhibit consistent trends.

The model residuals hover around zero, with an average MAPE of 12.4%, indicating reasonably accurate predictions.

4 Practical Applications

4.1 Acquisition Guidance

Because new users have long purchase cycles, their immediate value lags behind their lifetime value, making it unsuitable for short‑term spend decisions. With accurate LTV forecasts, ROI (LTV/CAC) can be set (typically >1) to determine the maximum allowable CAC, ensuring cost control.

Moreover, acquisition volume alone is insufficient; user quality matters. LTV helps allocate marketing resources toward high‑quality channels, improving overall ROI.

4.2 Fine‑Grained Operations

Even within the same SKU category, user value varies across demographics (region, occupation, age, interests, etc.). By breaking down LTV by these dimensions, businesses can diagnose differences, tailor personalized recommendations, and devise differentiated operational strategies to maximize revenue from a limited user base.

4.3 Product Iteration

Product updates are usually evaluated via conversion and retention metrics, which often correlate with user value. However, cases exist where conversion/retention remain stable while LTV rises (e.g., higher‑priced SKUs). LTV thus serves as a crucial metric to validate the business impact of product changes and guide investment decisions.

5 Outlook

Current LTV estimation in Liulishuo relies on a multi‑branch tree model differentiated by SKU type. Future work may incorporate user attribute dimensions alongside SKU types as model inputs, enabling more nuanced predictions that reflect attribute‑driven value differences.

6 References

[1] Hwang, H., Jung, T., & Suh, E. (2004). An LTV model and customer segmentation based on customer value: a case study on the wireless telecommunication industry. Expert Systems with Applications , 26(2), 181‑188.