The Three Ways to Pay for Advertising, and What Each One Hides

Every explainer on this subject gets the definitions right and the point wrong. Yes: CPM is cost per thousand impressions, CPC is cost per click, CPA is cost per acquisition. What the glossaries never say is that these are not three metrics measuring the same thing at different depths. They are three different contracts about who carries the risk when an ad does nothing, and once you read them as risk contracts, a lot of confusing pricing behavior across the industry becomes legible in an afternoon.

Start with CPM, the oldest deal in media. You pay for exposure, delivered or not to a human, noticed or not by anyone, and everything after the impression is your problem. The seller's risk ends the moment the ad renders. That sounds like a bad deal for buyers, and for undisciplined buyers it is, which is why the impression's defects, the half-viewable standard, the learned blindness, the wild variance in what an "impression" even certifies across channels, deserve the scrutiny I gave them in impressions are not created equal. A meaningful share of CPM dollars also evaporates before any human is even in the room, the supply-chain leakage I quantified in a third of your programmatic budget never reaches an impression. But CPM is also the cheapest possible way to buy attention when you know what you are doing, precisely because the seller has priced in none of your downstream success. Skilled brand advertisers live here on purpose: when your creative is strong and your measurement is honest, buying raw exposure wholesale beats paying someone else to guarantee outcomes at retail.

CPC moves one link down the chain. The seller now carries the risk of indifference; if nobody clicks, nobody pays. This is the deal that built Google, and it is a genuinely elegant alignment for demand capture: the platform only earns when a human acts. But look at what the contract incentivizes. A system paid per click is optimized, relentlessly, to generate clicks, which is not identical to generating customers. Curiosity clicks, misclicks, comparison shoppers, the click is the product being sold, and everything after it, the landing page, the offer, the close, returns to being your risk. The platforms know where their contract ends, which is why the improvements that pay best in CPC channels are usually on your side of the line, the territory I mapped in your landing page is killing more campaigns than your media.

CPA looks like the buyer's paradise: pay only when someone converts. The seller carries risk all the way to the outcome. Read the fine print of that deal, though, whether it is an affiliate arrangement, a lead-gen network, or a platform's target-CPA bidding, and two hidden clauses appear. First, whoever guarantees the outcome gets to define the outcome, and definitional creativity is the affiliate industry's core competency; a "conversion" can be a form fill from a sweepstakes page, an install that never opens, a lead that answers no phone, the quality collapse I detailed in cheap leads are the most expensive thing you can buy. Second, a seller carrying full risk prices that risk in, and steers volume toward the easiest conversions available, which are usually the people you would have won anyway. CPA deals do not eliminate risk. They convert performance risk into definition risk and incrementality risk, which are harder to see on an invoice.

Here is the practical geometry. These models form a slider between control and certainty: CPM gives maximum control and zero guarantees, CPA gives maximum guarantees and minimum control, CPC sits between. Where you should sit depends on which side of the deal you are better at than the market. If your creative and measurement are stronger than average, slide toward CPM and keep the margin the guarantors would have charged; this is what sophisticated teams are doing when they buy reach on raw CPMs and verify with the holdout methods from you do not need a data science team to run a holdout. If you are early, unmeasured, or thin on creative, sliding toward CPC and CPA rents competence you do not yet have, at prices that reflect it. And whatever the invoice says, compute all three numbers for every channel anyway, spend over impressions, over clicks, over acquisitions, because the pricing model is just where the receipt stops. The arithmetic should never stop there, and the moment a channel resists you calculating its effective CPA from first principles, you have learned the most important thing its pricing model was hiding.

A worked contrast makes the risk-contract lens concrete. A client once ran the same retargeting audience two ways in parallel: through a CPA network guaranteeing $38 per lead, and through direct CPM buying of the identical exchange inventory at a $4.10 CPM. The network delivered its leads at exactly $38, reliably, month after month. The direct buy, after our own optimization, produced equivalent-quality leads at a computed $22, because the $16 difference was the price of the guarantee, the network's margin for carrying risk we were fully equipped to carry ourselves. The reverse decision is just as real: for a founder with no measurement stack and no creative pipeline, that same $38 guarantee is a bargain against the $60 their inexperience would produce. The models are not good or bad. They are quotes for risk transfer, and the only question is which side of the transfer you should be on this year.

One closing translation, since the query that brought you here is usually asked by someone building their first serious plan: brand and reach objectives generally buy best on CPM, demand capture on CPC, and pure-risk experiments on CPA, with every arrangement graded, quarterly, in the currency that actually matters, which is incremental customers per dollar. The pricing models are how you pay. They were never how you should keep score.