Every marketing budget eventually poses the same question: should the next thousand euros go to the platforms, or to the customers? Media buys attention from people who have never heard of you. A reward budget buys a specific action from people who have already arrived. Both are performance spend, both are measurable, and most teams compare them on a number that makes the comparison meaningless.
This guide covers what each budget actually buys, why their cost-per-action figures are not comparable as reported, how the split should shift as traffic changes, and the case where paying customers is clearly the wrong move.
They sit at different points, so the raw numbers do not compare
Put a media CPA next to a cost-per-reward and the reward almost always looks cheaper. It usually is cheaper, and the comparison is still wrong.
Media spend is paying to create reach — to put you in front of someone who did not know you existed. Reward spend is paying to convert or retain reach you already have. The reward never runs unless somebody was already on your site, in your store, or watching your video. So the reward's cost per action excludes the entire cost of getting that person there, which the media line paid for.
Normalise before you compare. If you want the two on one axis, load the acquisition cost of the traffic into the reward's denominator, or compare them only on the job they do rather than on headline CPA. Otherwise you will conclude that rewards are five times more efficient than ads and quietly defund the thing that supplies the audience.
What each budget leaves behind
This is the difference that matters more than price, and it rarely appears in a media plan.
Media leaves nothing. When the campaign stops, the reach stops. You rent access to an audience, and at the end of the rental you own no more than you did at the start — some brand memory, and whatever you captured while they were with you.
A reward leaves a customer you can name. To be paid, someone has to identify themselves. You end up with a verified contact, a consent record, a dated action tied to a specific piece of content, and a row in your own database. That asset does not expire when the budget does, and it is what makes the next campaign cheaper.
That is the real argument for the reward line, and it is an argument about owning versus renting, not about CPA. The mechanics of paying for pre-purchase actions are in what social cashback is; how the payout itself is priced is in the cashback marketing guide.
The auction moves, your reward does not
Media prices are set by an auction you do not control. Costs rise when competitors bid, and they rise predictably in the fourth quarter — the same creative bought at the same target delivers less, and your plan needs headroom for it.
A reward is a price you set. It does not inflate because a competitor raised their budget, and it can be capped per customer, per period and in aggregate. In a quarter when media gets expensive, the reward line is the part of the budget whose unit economics still behave.
The flip side is equally important: rewards do not scale past your traffic. Double the reward budget with the same number of visitors and you mostly pay more to the same people. Media has a ceiling set by money; rewards have a ceiling set by audience size. That single asymmetry decides most allocation questions.
How the split should move
The right mix is a function of how much traffic you already have, not of preference.
Little or no audience. Almost everything to media, or to creators making content that earns distribution. A reward programme on a thousand visitors a month pays out a rounding error and teaches you nothing, because the sample is too small to read.
Meaningful traffic, weak conversion or weak repeat. This is where the reward line earns its place. You are paying to convert attention you have already bought, and every payout hands you an identified customer. Shifting budget here is usually the highest-return move available, because the expensive part — getting them there — is already paid for.
Plateau: traffic is fine and repeat purchase is fine. Rewards shift from acquisition to retention, and the question stops being cost per action and becomes cost per retained customer. That number is only trustworthy if you measure it against a holdout, for the reasons in our guide to measuring retention honestly.
Budget behaviour nobody warns you about
Reward spend is variable and lumpy. It fires only on qualifying actions, so it cannot overspend the way a media campaign can — but it also cannot be forecast like a fixed buy. Cap it per period, or a good month becomes an unexpected invoice.
Some rewards are never claimed. Any programme with a claim step has a share that goes unredeemed. That makes the accrued cost higher than the paid cost, and if you budget on paid-out figures you will under-reserve. Budget on accrual.
Part of the spend goes to behaviour you would have got anyway. Some share of any reward budget pays for actions that needed no incentive. Knowing that proportion is what turns cost per action into cost per incremental action, and it is the only version of the number worth defending in a budget meeting.
When paying customers is the wrong move
If nobody knows you exist, do not move budget out of media. Rewards multiply an audience; they cannot create one, and dividing a reward budget by no traffic gives you nothing.
If you cannot verify the action, do not pay for it. A reward attached to something self-reported and unverifiable is a fraud budget with a marketing label. Verification has to exist before the payout does.
If you have no way to store consent properly, hold off. The asset a reward budget produces is a contactable, lawfully-collected customer record; without that, you have paid for an action and kept nothing, which is the media outcome at a reward price.
And in a genuinely one-purchase category, the retention half of the argument disappears. What is left is a conversion incentive competing directly with a discount, and a discount is simpler.
The short version
Do not ask which is cheaper. Ask which constraint you are against. Short of audience, buy reach. Short of conversion or repeat among an audience you already have, buy actions — and make sure each payout leaves behind a customer you can contact again, because that residue, not the CPA, is the reason the line exists.
For how the reward mechanic is priced, capped and paid out, see TikJoy's Cashback API.