MyAIDevelopment · Product · Operations notes

Marketing Metrics · Unit Economics

The purpose of marketing metrics is not reporting but answering "Does it pay to buy more customers this way?"

Metric Tree

flowchart TD
    R[Revenue Growth] --> N[New Customers]
    R --> K[Retained Revenue]
    N --> V[Visitors or Leads]
    N --> C[Conversion Rate]
    K --> RT[Retention]
    K --> EX[Expansion]
    N --> CAC[CAC]
    K --> LTV[LTV]
    CAC --> U[Unit Economics]
    LTV --> U

Core Definitions

Metric Definition Caution
CAC Customer acquisition cost = (marketing + sales cost) / new customers State which costs are included
Paid CAC Paid channel cost / customers acquired through paid channels Distinguish from blended CAC
LTV Present value of net profit (or gross profit) over the customer relationship Overstated if calculated on revenue
LTV:CAC LTV / CAC Very sensitive to assumptions
CAC Payback Months needed to recover CAC State whether gross margin is applied
Retention Share of customers (or revenue) remaining after a period Distinguish logo-based from revenue-based
NRR Retention and expansion of existing-customer revenue (including expansion) Can be driven by one or two large customers
Conversion Rate Rate of conversion between stages Not comparable if the denominator changes

Blended CAC vs Paid CAC

a16z's "16 Startup Metrics" points out that if you only look at blended CAC, which includes customers who arrived organically, you cannot tell whether paid channels are actually profitable. To decide whether to increase budget, look at paid CAC separately.

Calculating CAC Payback

CAC Payback (months)
= CAC / (monthly revenue per customer × gross margin)

Example: CAC 6,000,000 KRW, monthly revenue 500,000 KRW, gross margin 80%
= 6,000,000 / (500,000 × 0.8) = 15 months

Calculating without gross margin makes the payback period look shorter than it really is.

Example: Unit Economics of the Incident-Monitoring Tool (Assumptions)

Using the same numbers as the payback example above, calculate LTV, LTV:CAC, and blended versus paid CAC in one pass. Every number is an assumption.

Assumptions
- ARPA (monthly revenue per account): KRW 500,000
- Gross margin: 80%
- Monthly logo churn: 3%, constant over time
- No expansion, no discounting to present value
- CAC (marketing + sales, blended): KRW 6,000,000

Monthly gross profit = 500,000 × 0.8 = KRW 400,000
Average lifetime     ≈ 1 / 0.03 ≈ 33.3 months
LTV                  ≈ 400,000 / 0.03 ≈ KRW 13.33M
LTV:CAC              ≈ 13.33M / 6.0M ≈ 2.2
CAC Payback          = 6.0M / 400,000 = 15 months

Splitting blended and paid CAC changes the picture.

New paying accounts this quarter: 10; total marketing + sales cost KRW 60M
- 4 via paid channels, paid-channel cost (ad spend + related sales cost) KRW 40M
- 6 via organic search and referrals, content and community cost KRW 20M

Blended CAC = 60M / 10 = KRW 6.0M  → LTV:CAC ≈ 2.2, payback 15 months
Paid CAC    = 40M / 4  = KRW 10.0M → LTV:CAC ≈ 1.3, payback 25 months

The paid-channel cost of KRW 40M includes, on top of the search-ad media spend in document 04's channel mix (KRW 2M a month, about KRW 6M a quarter), sales costs such as demos and onboarding support plus allocated staff cost (assumption). Dividing media spend alone makes paid CAC look far lower than it is. The KRW 20M on the content and community side is likewise that channel's media spend in document 04 (about KRW 6M a quarter) plus production staff cost, and the lifecycle and test budgets (about KRW 3M a quarter) are counted on this side too (assumption).

Looking only at the blended number, adding paid spend seems fine; isolate the paid channels and payback takes more than two years. Decide whether to raise the budget using paid CAC (ideally an incremental CAC estimated as in document 06). At this scale a user-level lift experiment is underpowered, so estimate it with an on/off or geo test as documents 06 and 04 describe. If churn falls over time or there is expansion, LTV grows, so recalculate whenever an assumption changes.

Benchmarks Are Only a Reference

Frequently cited rules of thumb:

These numbers vary with industry, pricing model, growth stage, and cost of capital. Your own cohort data is the more important benchmark.

Look at Cohorts

Averages are good at lying. Split by signup month and acquisition channel.

            M0    M1    M2    M3
Jan cohort  100%  62%   51%   47%
Feb cohort  100%  58%   45%   40%
Mar cohort  100%  66%   57%   -

Reading Funnel Metrics

Stage Example metrics Common misconception
Awareness Impressions, reach, brand search volume More impressions means better results
Acquisition Visits, signups, leads More leads is always good
Activation Rate of reaching the first value experience Signup = active user
Revenue Paid conversion, ARPU A conversion bought with discounts counts the same
Retention Retention rate, churn rate The overall average is enough
Referral Referrals and invites A referral made for a reward counts the same

Bad Example / Good Example

Bad example:

Leads grew 40% this quarter.

Good example:

Paid-channel leads grew 40% this quarter, but the paid conversion rate fell, so paid CAC rose 12%. Three-month retention of the new cohort is unchanged, so we keep our LTV assumption. Next quarter we will cut budget for the two campaigns with the lowest conversion rates and run landing page experiments.

A good report flows from number → interpretation → decision.

Metric Pitfalls

References