Beta
EconomyReading · ~3 min · 65 words deep

LTV / CAC

LTV/CAC compares lifetime revenue per customer to acquisition cost · >3× is the SaaS healthy benchmark · AI consumer apps are 2-4× so far.

TL;DR

LTV/CAC compares lifetime revenue per customer to acquisition cost · >3× is the SaaS healthy benchmark · AI consumer apps are 2-4× so far.

Level 1

LTV/CAC = (average revenue per customer over their lifetime) / (cost to acquire that customer). SaaS benchmark: >3× is healthy, >5× is excellent. Consumer apps are lower due to higher churn. AI apps are still proving their LTV · Perplexity reportedly at ~2.5×, ChatGPT Plus at 2-3× (low because of usage-cost headwinds). Early AI numbers are noisy but improving.

Level 2

LTV calculation is fragile · typical formula: (average monthly revenue × gross margin) / monthly churn rate. For ChatGPT Plus at $20/month, 60% gross margin, 5% monthly churn, LTV = $240. If CAC is $80, LTV/CAC = 3×. Higher-churn apps (weekly AI image generators) run lower. Enterprise AI is different · LTV extends 3-5 years with NRR >100%, CAC is heavy but ratio often 5-10×.

Level 3

LTV/CAC for AI has complications. Inference costs can scale with usage (heavy users subsidize light users), distorting "LTV per customer." Some AI apps now structure pricing to cap downside · credit-based (Midjourney, Runway) or usage-tiered (Claude Pro). Measuring CAC cleanly requires separating paid marketing from organic (word-of-mouth dominates AI). Best practice: track LTV/CAC cohort by cohort and adjust for compute cost changes over time.

The takeaway for you
If you are a
Researcher
  • ·LTV = (ARPU × GM) / churn rate
  • ·Healthy ratio: 3× · great: 5×
  • ·AI consumer apps: 2-4× · enterprise: 5-10×
If you are a
Builder
  • ·Cohort analysis > average
  • ·Watch CAC inflation as paid channels crowd
  • ·AI app churn can be deceiving · measure by engagement, not signups
If you are a
Investor
  • ·Primary SaaS-era metric · applied unevenly to AI
  • ·Low LTV/CAC in AI is often temporary · improves as retention matures
  • ·Track enterprise-only cut separately · much cleaner signal
If you are a
Curious · Normie
  • ·Does a customer spend more than it costs to get them? >3× = yes, profitable
  • ·Healthy tech businesses score 3× or higher
  • ·AI apps still figuring out their numbers
Gecko's take

AI LTV/CAC is the hardest metric to trust right now · too many models in flux. Anchor on enterprise cohorts, not blended.

ARPU × gross margin / monthly churn. Use 60-70% GM for API-heavy AI apps.