LTV by acquisition channel, and why your best channel might be your worst

Most acquisition decisions are made on the cost of getting a customer. The more useful question is what that customer is worth afterwards, because the two vary independently — and the channel that looks cheapest at signup is frequently the one whose customers leave first.

The blended number hides the problem

The standard lifetime value formula:

LTV = (average revenue per account × gross margin) ÷ churn rate

Run it across all your customers and you get a single figure that describes nobody. Consider a business with $450 blended LTV:

Channel	ARPA	Monthly churn	LTV (at 90% margin)	CAC	Ratio
Comparison content	$20	2%	$900	$120	7.5:1
Newsletter sponsorship	$20	4%	$450	$150	3.0:1
Discount campaign	$20	9%	$200	$90	2.2:1

Same price, same product, same blended $450. The discount campaign has the lowest acquisition cost and the worst return; the comparison content costs more per customer and is worth two and a half times as much.

A CAC-only dashboard recommends the discount campaign. It is the wrong recommendation, and nothing in the blended numbers reveals it.

Why channels differ this much

The mechanism is selection, not marketing quality.

Intent at arrival. Someone who read a detailed comparison before clicking has already decided what problem they are solving. Someone who clicked a discount is often buying the discount, and churns when it ends.

Fit. Content that explains a specific use case attracts people with that use case. Broad-reach advertising attracts everyone, including people the product does not suit.

Expectation. A channel that oversells produces customers whose first week is a disappointment. That shows up as month-two churn, months after the campaign was declared a success.

None of this is fixable by improving onboarding for the discount cohort. The channel selected for a different kind of person.

How to compute it

Step 1: tag the acquisition source and keep it. The channel a customer arrived from has to survive into your customer record. In practice this means capturing the source at signup and storing it on the account — UTM parameters for tagged traffic, the referrer for the rest.

Note the constraint honestly: this records the source of the visit that converted. If the customer first heard of you six weeks earlier, that touch is not in the record unless you are keeping cross-day visitor identity, which is a real privacy trade and one you should make deliberately.

Step 2: cohort by channel, not by month. The rows are channels, the columns are months since signup, and the cell is the share still paying. That is a retention cohort grid with the channel as the grouping.

Step 3: read the curve shape, not the starting point. Every curve falls. The question is whether it flattens. A channel whose curve plateaus at 60% has found people the product suits; one that keeps falling produces customers who all leave eventually, and its LTV is bounded no matter how cheap the clicks are.

Step 4: divide by CAC. LTV alone is a number. LTV over acquisition cost is a decision.

What to do with the answer

Reallocate towards the ratio, not the cost. If comparison content returns 7.5:1 and paid social returns 2.2:1, the marginal budget goes to the first even though each customer costs more.

Check the payback period before celebrating. A 7.5:1 ratio with a 30-month payback can still put you out of business if cash is finite. Payback is CAC divided by monthly gross profit per customer, and under twelve months is the usual comfort line for self-serve software.

Do not kill the discovery channel. A channel with poor last-click LTV may be introducing customers who convert through another channel later. Check it under first-click before cutting it — this is the most common way a working top-of-funnel channel gets defunded.

Fix retention where the volume is. If your largest channel has middling retention, a small improvement there outweighs a large improvement in a small channel. Multiply before prioritising.

When not to bother

If you have forty customers, this analysis will produce numbers and none of them will mean anything. One cancellation in a channel with six customers moves that channel’s churn rate by seventeen points.

Below roughly a few dozen customers per channel and three months of history, do the simpler thing: look at revenue per visitor by channel, which needs far less data and answers a related question. Come back to LTV when the cohorts are large enough to survive a single cancellation.

The uncomfortable version

Sometimes the channel that fails this analysis is the one you are best at. A founder who is good at a particular kind of campaign will find reasons why the retention data is misleading.

The retention data is usually not misleading. It is measuring something the acquisition metrics structurally cannot see, which is what happens to the customer after the part you were optimising.

---

sonex reads revenue from Stripe or Polar and reports retention cohorts and channel attribution beside your traffic, on plans that start free. Start free.