Target Cost per Acquisition: How to Set One You Can Defend

Most target CPAs are set one of three ways: last year's actual with ten percent shaved off it, a figure a competitor mentioned on a podcast, or the number that makes the plan work when you solve the spreadsheet backwards from the growth target. All three produce a number nobody can defend when it is missed, because none of them started from anything the business actually knows about itself.

A target you can defend starts from gross margin and a payback tolerance, and it ends as a set of allowable costs at every stage of the funnel, so the person buying media or booking meetings knows their own ceiling without having to reason about lifetime value.

Start from gross profit, and from cash

Revenue is the wrong starting point. A dollar of revenue at 40 percent margin buys you less acquisition than a dollar at 85 percent, and a target CPA built on revenue silently assumes margins you may not have.

The two inputs that set the ceiling are gross profit per customer and how long you are willing to wait to get it back.

Gross profit per customer is annual contract value multiplied by gross margin, with margin computed properly: hosting, support, delivery headcount, payment processing, any third party fee that scales with usage. For a transactional business, substitute gross profit per order multiplied by the number of orders you can honestly expect inside a defined window, using the repeat behaviour you have already observed rather than the one in the plan.

Payback tolerance is the number of months of that gross profit you are willing to spend to win the customer. This is a cash decision before it is a marketing decision. A company with two years of runway and a company with seven months can look at identical unit economics and correctly choose different tolerances. Anyone with meaningful working capital constraints should set tolerance from the cash they can be out of pocket at peak, not from a benchmark for their category.

Allowable acquisition cost is those two multiplied together. Monthly gross profit times the number of months of payback you will accept. That is the ceiling, and everything downstream is a division of it.

1. Step 1Compute gross profit per customer

   Annual contract value times true gross margin. Include delivery, support and usage-based third party costs.

2. Step 2Choose a payback tolerance

   How many months of gross profit you will spend to win a customer. Set it from your cash position, not from a category benchmark.

3. Step 3Multiply for allowable CAC

   Monthly gross profit times payback months. This is the total ceiling for everything acquisition costs, including the people who close.

4. Step 4Subtract the cost of closing

   Loaded sales cost per new customer comes out first. What remains is the budget available to generate the pipeline.

5. Step 5Cascade with your own rates

   Multiply down the funnel by your observed conversion rates to get an allowable cost per opportunity, per meeting and per qualified lead.

Subtract the cost of closing before you cascade

The step almost every model skips: allowable CAC covers every cost of acquiring the customer, and closing is one of those costs. The account executive's loaded salary, their commission, the sales engineer on the technical call, the CRM seats. If you cascade the full allowable down the funnel, you will authorise a cost per meeting that leaves nothing to pay the person who takes the meeting.

Compute loaded sales cost per new customer, subtract it from allowable CAC, and cascade what remains. That residual is the honest demand generation budget per customer, and it is usually a sobering fraction of the headline number.

A worked illustration with chosen inputs

Everything in this section uses invented figures, picked because they divide cleanly. They are an illustration of the method rather than typical values, and substituting your own numbers is the entire point of the exercise.

Assume a business selling a $30,000 annual contract at 75 percent gross margin. Annual gross profit per customer is $22,500, so monthly gross profit is $1,875.

Assume the company chooses a 12 month payback tolerance, based on its own cash position. Allowable CAC is 12 times $1,875, which is $22,500.

Assume loaded sales cost, meaning the closing team's salary, commission and tooling divided by new customers won, comes to $7,500 per new customer. Subtracting it leaves $15,000 of allowable demand generation cost per new customer.

Now cascade using the company's own observed conversion rates. Assume 20 percent of qualified opportunities become customers, 40 percent of held meetings become qualified opportunities, 75 percent of booked meetings are actually held, and 30 percent of qualified leads become booked meetings.

Twenty percent win rate means each opportunity is worth one fifth of a customer, so the allowable cost per opportunity is 20 percent of $15,000, which is $3,000. Forty percent of held meetings becoming opportunities gives an allowable of $1,200 per held meeting. A 75 percent show rate gives $900 per booked meeting. Thirty percent of qualified leads converting to booked meetings gives $270 per qualified lead.

Allowable CAC$22,500

12 months of gross profit at a $30,000 contract and 75% margin

Available for demand generation$15,000

After subtracting $7,500 of loaded closing cost per new customer

Per qualified opportunity$3,000

At an assumed 20% win rate

Per held meeting$1,200

At an assumed 40% meeting to opportunity rate

Per booked meeting$900

At an assumed 75% show rate

Per qualified lead$270

At an assumed 30% lead to booked meeting rate

That last figure is the one that changes behaviour. A team that knows its allowable is $270 per qualified lead can evaluate a channel, a list source or a supplier quote in about ten seconds, without escalating and without reasoning about lifetime value in the meeting.

Where the model breaks

Conversion rates from small samples. A 20 percent win rate computed from fifteen opportunities is an anecdote with a percent sign attached. The cascade multiplies four such estimates together, and the error compounds. Where the sample is thin, use a wide band rather than a point estimate, plan against the pessimistic end, and treat the optimistic end as upside rather than as the budget.

Stage definitions that mean different things to different people. The cascade only works if a qualified opportunity means the same thing every time it is counted. Loose stage definitions inflate the middle of the funnel, which inflates the allowable at every stage below it, which authorises overspending on exactly the leads that were never going to convert. The distinction between an MQL and an SQL is not a taxonomy argument when a budget ceiling is derived from it.

Churn hiding inside the margin. Payback tolerance is only safe if customers survive it. A twelve month payback against a product with heavy first year churn means a meaningful share of customers never repay their acquisition cost at all. Where retention is weak, either shorten the tolerance or compute against the gross profit actually realised in the first year rather than the contracted amount.

Mix. One target CPA across segments that convert differently will overpay in the weak segment and underpay in the strong one. If enterprise wins at half the rate of mid market at three times the contract value, those are two targets. Set them separately, which usually means writing the ideal customer profile precisely enough that a lead can be assigned to a segment before it is priced.

Average against marginal. The allowable is an average ceiling. The cost of the next unit of volume in a channel is typically higher than the cost of the last, because the most responsive audience is reached first. A channel sitting exactly at the allowable is already unprofitable at the margin.

Making the target defensible in a room

A target survives challenge when the challenger can see which input they disagree with. Publish the target with its inputs attached: margin, payback tolerance, loaded closing cost, and the four conversion rates with the sample size behind each. Then a disagreement becomes an argument about a specific rate rather than about whether the number feels right.

Two habits matter as much as the arithmetic. Review on a fixed cadence, quarterly for most companies, and change the target only when an input has actually moved rather than because performance was disappointing. Moving the target to match the result is how a target stops meaning anything. And record the change with its reason when it does move, so a target that has crept upward over four quarters is visible as a trend rather than as four separate reasonable decisions.

Is this target defensible?

• ✓Yes: It derives from gross profit, not revenue
• ✓Yes: Payback tolerance was chosen from the cash position and written down
• ✓Yes: Loaded closing cost was subtracted before cascading
• ✓Yes: Each conversion rate has a sample size recorded next to it
• ✓Yes: Separate targets exist for segments that convert differently
• ✓Yes: Stage definitions are written and applied consistently
• ~Depends: First year churn is reflected in the payback tolerance
• ✓Yes: The target moves only when an input moves
• ✕No: The target was set by working backwards from the growth plan
• ✕No: The target came from a benchmark for the category

What to do when you keep missing it

A target missed consistently is information, and the useful first question is which stage is missing it. The cascade makes that answerable, because each stage has its own allowable and its own actual. Overspending at the meeting stage while converting above plan at the opportunity stage is a different problem from the reverse, and they have opposite remedies.

If the miss is concentrated at the top, the pool or the targeting is wrong, and buying more of the same audience will not fix it. If the miss is in show rate or in meeting quality, the definition of a qualified meeting is doing less work than it should. Where meetings are bought from a supplier, the definition needs to be agreed in writing before anything launches, covering audience, buyer responsibility, the prospect agreeing to a relevant business conversation, attendance, and exclusion of accounts already disclosed as customers or open opportunities. Budget, timing and decision authority do not belong in that definition, because they describe how the conversation goes rather than whether the right conversation happened. Our breakdown of pay per appointment pricing in B2B covers what that definition has to contain to be enforceable, and outsourced SDR pricing covers how the same ceiling compares against a headcount model.

If the miss is at the bottom, the win rate assumption was wrong, and the honest response is to lower the allowable rather than to keep spending against a rate the pipeline is not producing.

Whichever it is, the target has done its job by localising the problem. A single blended CPA number can only tell you that something is expensive.

If you would rather test the cascade against real numbers from your own market before committing budget to it, you can see what a campaign would look like.