performetra

How to price a CPA offer without losing money

Most unprofitable performance campaigns were unprofitable on the day the payout was agreed. Here is the arithmetic that prevents it.

TL;DR — the short version

Rule 1
Price payouts from contribution margin, never from revenue
Rule 2
Translate allowable CPA through funnel rates to the priced event
Rule 3
Discount for returns, refunds and rejections before signing
Rule 4
Leave enough margin for publishers, or the offer sits unpromoted

Most unprofitable performance campaigns were unprofitable on the day the payout was agreed. The traffic was fine, the tracking worked, and the network delivered what it promised. The number at the top was simply wrong, and every conversion after that made the problem larger.

Setting a payout is arithmetic, not negotiation. Here is the version we use.

Start from contribution margin, never revenue

The most common mistake is pricing a payout as a percentage of revenue. Revenue pays for the product, the shipping, the payment gateway, the returns and the support ticket. What is left is contribution margin, and that is the only pot an acquisition payout can come from.

Take an order with an average value of ₹2,400. Cost of goods at 55% leaves ₹1,080. Shipping and packaging at ₹120 leaves ₹960. Payment gateway and platform fees at 2.5% take another ₹60. Support and reverse-logistics provisioning, say ₹80. Contribution margin is around ₹820, not ₹2,400.

An 8% commission sounds modest against revenue. Against contribution margin it is 23%. That may still be a good deal — but it is a different conversation, and it is the honest one.

Decide what share of margin acquisition can take

There is no universal answer, but there is a useful frame: how long can you wait to make money back?

  • Single-purchase products. Acquisition has to come out of the first order’s margin. Most brands can sustain 30–50% of contribution margin, and above that the business runs on hope.
  • Repeat-purchase products. You can spend the first order’s entire margin, and sometimes more, provided you actually know your repeat rate rather than assuming one.
  • Subscriptions. Payback period is the real constraint. A twelve-month payback is a financing decision, not a marketing one, and it needs whoever controls the cash to agree.

Work backwards through the funnel

Once you have an allowable cost per acquired customer, translate it into the event the publisher is actually paid on. A lead is not a customer, and an install is not a user.

Suppose your allowable is ₹400 per acquired customer and 20% of validated leads become customers. Your maximum CPL is ₹80 — before any lead rejection. If 10% of leads are rejected on validation, the payable CPL drops to about ₹72 to hold the same economics.

Run that calculation for every step you price at, and the difference between paying for an install and paying for a registration stops being a preference and becomes a number.

The further upstream you price, the more of the funnel risk you are buying. That can be the right call — it usually gets you more volume — but it should be a decision, not an accident.

Price in the reversals before they happen

Three reversal types quietly break payouts:

  • Returns and cancellations. In cash-on-delivery markets, 25–40% of placed orders may never complete. If commission is paid on placed orders, that entire failure rate is a pure loss.
  • Refunds inside a trial window. Subscription trials that cancel on day six cost you the acquisition payout and produce no revenue at all.
  • Lead rejections. Even validated leads get rejected. Whether that rejection is honoured after invoicing is a contract term, and it should be a written one.

The clean fix is to move the billable event past the reversal risk: pay on delivered orders, on paid conversions, on contacted leads. Where that is genuinely not measurable, discount the payout by the historical reversal rate and put a clawback window in the insertion order.

Leave room for the network to work

A payout that leaves nothing for the publisher after their media cost will not deliver volume, no matter how enthusiastically it was agreed. Publishers buy traffic at a market price; if your payout does not clear that price plus their margin, your offer sits in the platform unpromoted while everyone politely blames the creative.

Before finalising, ask the network what comparable offers in your category pay. If your number is 40% below the market, you have not saved money — you have bought silence.

Three checks before you sign

  • Model the scaled month, not the test. Test-month economics benefit from your best publishers. At scale the mix gets worse. If the plan only works at test-level quality, it does not work.
  • Set the cap where a mistake is survivable. Caps are cheap. Overdelivery on a mispriced offer is not.
  • Agree the reconciliation date up front. A payout with no reconciliation process is a payout you cannot correct.

None of this is complicated. It is simply done at the wrong time — usually after the first invoice, when the arithmetic becomes everyone’s problem at once.


Written by the Performetra campaign team. If you want this applied to a live campaign rather than read about, tell us what you are running.

Questions this article answers

The fixed amount an advertiser pays per conversion — install, lead, action or sale. It is the core price in performance marketing, and getting it wrong makes every subsequent conversion a small loss.

Because revenue also pays for the product, shipping, gateway fees, returns and support. Only contribution margin is available to buy customers with; an 8% commission on revenue can quietly be 23% of margin.

Single-purchase products: typically 30–50% of contribution margin. Repeat-purchase products can spend the whole first-order margin if the repeat rate is measured, not assumed. Subscriptions are constrained by payback period.

Multiply through the funnel: if a customer is worth ₹400 in acquisition spend and 20% of validated leads convert, the maximum CPL is ₹80 — less after expected rejections.

Returns and COD failures, refunds inside trial windows, and lead rejections. Either move the billable event past them — delivered order, paid conversion — or discount the payout by the historical reversal rate.

The agreed period during which a billed conversion can be reversed — an order returned, a lead rejected. It belongs in the insertion order with its length and mechanism stated, typically 30 days.

Publishers do not promote it — their media cost does not clear. You have not saved money; you have bought silence. Ask the network what comparable offers pay before finalising.

Test months benefit from the best publishers; at scale the traffic mix gets worse. If the economics only work at test-level quality, they do not work.

Where a pricing mistake is survivable. Caps cost nothing; overdelivery on a mispriced offer is real money. Raise them on data, not optimism.

Yes — send the conversion event, geos and volume via the contact page and you get a written payout and expected volume band, usually the same day.

Next step

Turn this into a campaign.

We price outcomes across app growth, lead generation, e-commerce and creator marketing. Send us the event you want to pay for.