Historical LTV Analysis

Historical LTV (Life-Time Value) measures a cohort's cumulative average revenue per user on day N after registration, making it the core model for evaluating long-term user value and channel quality. It corresponds to the backend AnalyticsService.LTV.

1. What Questions It Answers

  • What is a cohort of new users' cumulative average payment (LTV) on day 7 / day 30 after registration?
  • Among users acquired from different channels, whose LTV curve is higher? (evaluating long-term channel quality)
  • Payback period: when does LTV exceed the customer acquisition cost (CAC)?

2. Key Concepts

ConceptMeaning
CohortThe group of users selected by register_time (registration time range)
day_nDay N after registration (0 = registration day)
LTVThe cohort's cumulative average payment up to day N (total payments / cohort size)
cohort_sizeNumber of users in the cohort

LTV is a curve that increases with day_n (a cumulative value that only grows).

3. Backend API

gRPC: AnalyticsService.LTV

FieldTypeDescription
time_rangeTimeRangeRegistration time range (filters the cohort by register_time)
app_iduint32 (optional)Filter by app
dimensionstring (optional)Attribution dimension, default channel (LTV grouped by channel)

Response LTVResponse: points[] (LTVPoint contains label dimension value / day_n / ltv / cohort_size / total_amount), max_days (maximum observed days).

HTTP (admin proxy)

POST /admin/v1/analytics/ltv
Content-Type: application/json

{
  "timeRange": { "startMs": 1717459200000, "endMs": 1718064000000 },
  "dimension": "channel"
}

4. Typical Scenarios

Long-term channel quality evaluation

dimension=channel: compare LTV curves across channels. One channel may have low CAC but also low LTV (low-quality traffic); another may have high CAC but far higher LTV (high-quality users worth investing more in).

Payback period estimation

The intersection of the LTV curve with CAC (customer acquisition cost) is the payback point. The faster LTV grows and the earlier payback happens, the healthier the channel.

Cohort value monitoring

Fix a registration cohort and keep watching whether its LTV curve grows as expected, evaluating the product's monetization capability.

5. Notes and Caveats

  • The time range is registration time: time_range filters register_time (to select the cohort), not payment time.
  • LTV is cumulative: day-N LTV includes all payments from day 0 to N, and only grows.
  • Depends on registration and payment events: both the register event (to define the cohort) and payment events with amount must be collected correctly.
  • Observation window limits: max_days is bounded by how long data has existed — a newly registered cohort only shows the first few days of LTV; long-term LTV requires data accumulation.
  • Difference from Revenue: Revenue Analysis looks at GMV in a time window; LTV looks at a specific cohort's long-term cumulative per-user value.

6. Related Documents