Customer Acquisition Audit · System 03 · deciding where the next dollar goes
Decide With Cohorts, Not Averages
“LTV:CAC should be at least 3x” gets repeated in enough board decks that it starts to sound like a law of business. It isn’t one. It’s a rule of thumb published by a handful of venture investors and SaaS benchmarking firms, built from portfolios of companies that don’t look like most small and midsize businesses, and it varies by a factor of two depending on which year’s report you’re reading.
The idea in one line: use payback and LTV:CAC as tools for comparing your own channels and cohorts to each other, built from numbers you actually observed. Treat any externally published threshold as a sanity check at most, never as the bar a channel has to clear to be considered healthy.
01
Heuristics, not lab results
The two most quoted numbers in acquisition economics come from operator experience, not from independent research designed to establish a universal standard.
CAC payback
How many months of gross-margin profit it takes to recover CAC. Commonly cited targets range from under 12 months for SMB SaaS to under 24 for enterprise — published mainly by venture firms and SaaS benchmarking surveys, and the “good” band has moved between their own reports.
LTV:CAC
Lifetime value divided by acquisition cost. “3x or better” is the most repeated version; the same publishers have used 4–6x for other segments. Academic customer-lifetime-value research supports the concept of comparing value to acquisition cost, not any specific ratio as a threshold.
None of this means the numbers are useless. It means a channel sitting at 2.4x LTV:CAC isn’t automatically broken, and a channel at 3.2x isn’t automatically healthy. What decides that is how it compares to your other channels, and whether the trend is improving or getting worse.
02
Use your own cohort’s realized numbers
A static formula run off a single ARPA and churn assumption is a reasonable dashboard number. It’s a weak basis for a real budget decision.
The cohort-based version
For each acquisition cohort, sum realized revenue × gross margin, month by month, until it equals or exceeds that cohort’s acquisition cost. The month that happens in is the realized payback.
This version naturally handles ramp periods, discounting, usage-based revenue, and early churn — all of which a static ARPA-over-churn-rate formula quietly assumes away. It takes longer to build once and is worth building once.
Do this per channel and per segment, not just for the business as a whole. A company-wide payback number can look perfectly healthy while hiding one channel that’s dragging it down and another that’s carrying it, which is exactly the pattern System 02 found in the home-services example.
03
Decide what changes, channel by channel
1
Grow it. Payback is fast, cohort margin is durable, and volume is still small relative to the market. The clearest case for more budget.
2
Hold it. Payback and retention are acceptable but not exceptional, and the channel is already near its practical volume ceiling.
3
Fix the qualification. CAC is reasonable but retention is weak. The channel is probably bringing in the wrong customers rather than being the wrong channel — tighten targeting or add a qualification step before cutting it.
4
Test before trusting it. For anything where attribution might be harvesting demand you already had — branded search chief among them — run the incrementality test in Where the Next Dollar Goes before making a budget call either way.
5
Cut it. Weak payback, weak retention, and no plausible fix to the qualification. The channel is buying customers the business would be better off not having.
04
A worked example
Back to the home-services company from System 02 and its four channels, now with a decision attached to each.
The four decisions
Cold outbound: grow it. Highest realized 180-day margin of the four, and volume is a fraction of what the sales team could handle.
Local service ads: hold it. Solid payback, already close to the local market’s volume ceiling — more budget here buys more expensive clicks, not more customers.
Referral program: fix the qualification. Cheap and high-volume, but margin is thin because too many referrals are one-off small jobs. Add a minimum-job-size filter to the referral incentive before deciding whether to grow or cut it.
Branded search: test before trusting it. Nobody else is bidding on the brand name today, which is exactly the condition under which the research says paid brand ads mostly substitute for organic clicks. Pause it in one region for four weeks and measure total branded traffic — organic included — before reallocating that budget.
Ranked by CAC alone, this company would have kept pouring budget into its cheapest channel and started worrying about its most expensive one. Ranked by realized cohort economics, the decision runs almost exactly the other way.
Systems 01 to 03 are one method in three passes: load the real cost, judge the customers it bought, then decide where the next dollar goes using your own numbers instead of a borrowed threshold. The complete system assembles them into something you can run start to finish on your own acquisition spend, with the evidence to gather first and a master prompt to paste into whatever assistant you already use.
