Try explaining CPM to someone outside marketing.
“Well, you pay for a thousand people to maybe see your ad. But you don’t know if they actually made a purchase. At least they saw it and it’s in their head, though, right?”Right?
Not the most convincing argument. But that’s effectively the deal advertisers have been selling (and brands have been accepting) for decades. Hand over a budget, someone shows your ad to a lot of people, and you hope the math works out downstream.
Meanwhile, banks and card networks can now confirm, transaction by transaction, whether someone actually bought something. So, if you can pay for a verified sale, why are you still paying for a guess?
Below, we dive more into the history of paid advertising and why a more modern “pay for performance” approach makes more sense with shoppers who, a majority of time, tune ads out, and when brands don’t have as much to spend.
How We Ended Up Buying “Eyeballs”
In print, you paid for circulation because there was no way to know who read page 34.
In broadcast, you paid for rated audience because a Nielsen panel was the closest thing to a measurement anyone could get.
Both were reasonable bets in a world with no other signal. You couldn't follow a magazine reader to the checkout counter, so you paid for the reach and trusted the rest.
Digital inherited that logic wholesale. When banner ads showed up in the mid-90s, the industry reached for the unit it already knew — impressions — and priced it per thousand, exactly like print.
The tooling, the dashboards, the agency commissions all got built on top of that assumption. What nobody stopped to ask was whether "someone might have seen this" was still the right thing to buy once you finally could see further down the funnel. The proxy outlived the reason for the proxy. That's most of the story of why CPM is still here.
The Case Against Paying for Impressions
An impression is not evidence anyone saw your ad. It's evidence your ad loaded on a page.
Bot traffic inflates impression counts without inflating real human attention. Below-the-fold placements and auto-refreshing inventory rack up “views” that never crossed anyone's field of vision.
The ad industry has known about this for years and mostly just adjusted the discourse around it (hello, viewable CPM) instead of fixing the underlying problem, which is that you're still buying a chance, not a result.
Pay-for-performance replaces that possibility with a receipt.
What Pay-for-Performance Advertising Means
Pay-for-performance advertising is a pricing model where the advertiser only pays when a defined outcome happens: a sale, a lead, a redemption. Not when an ad loads on a page or when someone (read: probably a bot) clicks.
Compare that to the two models most digital advertising still runs on:
- CPM (cost per mille) charges an advertiser per 1,000 impressions, regardless of what anyone does after seeing the ad.
- CPC (cost per click) charges per click, which gets you closer to intent but still stops short of a transaction. Someone can click an ad, get distracted by a text, and never come back.
Pay-for-performance skips both of those proxies so you’re paying for what you want: a real transaction.
Why CPM and CPC Persist Anyway
While it doesn’t make a ton of sense on paper, there are few key reasons why these metrics are still used everywhere today:
1. Programmatic ad buying was built around exposure metrics from the start
Agencies built commission structures and reporting dashboards around media spend, not results. TV-era brand safety habits (show up next to the right content, in front of the right eyeballs) carried straight into digital without anyone rethinking whether “eyeballs” was still the right unit to buy.
2. Publisher and ad networks don’t love a pay for performance model
Why? Because they only get paid when their inventory actually leads to a sale. That means they're betting their own revenue on their traffic converting.
If their audience is engaged, well-targeted, and likely to buy, that's a bet they'll happily take, because they'll get paid either way and possibly earn more per outcome.
If their audience is low-quality, poorly targeted, or largely bots and passive scrollers, taking on a performance model is a losing bet for them. Their inventory won't convert, so they won't get paid. So instead, they keep selling on CPM, where they get paid the moment the ad loads, whether or not anyone real ever saw it or bought anything.
A lot of networks would rather not make that bet, so they keep selling impressions and let the advertiser absorb the uncertainty instead.
3. Attribution is hard
It's tough to prove a sale came from an ad than to just count how many times the ad loaded. But for anyone with access to transaction-level data, that's not the obstacle it used to be.
Merchant-Funded Rewards are Pay-for-Performance in Its Best Form
When a brand runs a cash back offer through a commerce media network, the brand funds the reward. But the reward only pays out when a cardholder buys something. No purchase, no payout.
The "impression" and the "result" are the same event, so instead of paying up front to be seen and eating the risk that nothing happens, the brand pays on the back end, for outcomes it can verify.
And because the reward is tied to a real transaction on a real card, it lines up perfectly with where a shopper already is. A cardholder browsing offers in their banking or BNPL app isn't being interrupted, they’re getting a sweet deal.
It also gives brands levers a CPM buy never could.
Offers can be shaped around behavior: a bit more cash back on each repeat purchase to build a habit, a boosted rate a cardholder taps to activate, a seasonal push timed to when people are already spending in a category.
Read more about the perks of offering cash back rewards →
What Makes This Possible Now (That Wasn't Available 10 Years Ago)
Performance marketing as a category has existed for a while, but most of it still relies on self-reported or pixel-based "conversion" events. A form fill, a pixel firing on a thank-you page.
Those are easy to spoof, easy to double-count, and easy to lose in a browser that's blocking third-party cookies.
But banks and card networks can confirm a purchase actually happened. With commerce media, the “action” being paid for is a verified purchase, not a self-reported click funnel.d you know? Kard is the first independent commerce media network. Using predictive AI and first-party transaction data from millions of Gen Z and Millennial shoppers, Kard powers hyperpersonalized offers that scale customer acquisition.
CPA vs. CPM: What Changes When You Switch the Unit You're Buying
Say a brand spends $50,000 on a CPM campaign at a $10 CPM. That buys 5 million impressions. It's a great headline number, but it tells you almost nothing about what happened next. Maybe 40,000 people bought something, maybe 4,000.
Now put that same $50,000 into a pay-per-sale model. If the cost per sale is $50, that's 1,000 confirmed transactions. That’s that.
Critics of CPA will point out that the headline cost per action is usually higher than CPM's headline cost per impression. That's true, but it misses the point.
A $10 CPM that converts at 0.1% and a $10 CPM that converts at 2% cost exactly the same on the invoice, even though one is worth ten times the other. Under CPM, the advertiser absorbs the risk that a campaign flops while the publisher gets paid regardless.
Under a performance model, the people running the campaign only get paid when it works — which tends to make them a lot more invested in making it work.
What to Ask Before You Move Budget to a Performance Model
Not every "performance" pitch means the same thing, so a few questions are worth asking before you sign anything.
How is Performance Defined and Verified?
A redemption is not the same thing as a purchase, and a purchase confirmed by a pixel is not the same thing as one confirmed by a bank. Ask exactly what triggers payment.
Who Owns the Risk If the Campaign Underperforms?
If the answer is "you still pay regardless," you haven't actually left the CPM model, you've just renamed it.
Can the Partner Show Incrementality, Not Just Attribution?
Vendors who can only show you attribution (this ad ran, this sale happened, therefore…) haven't proven anything a coincidence couldn't also explain.
Look for partners who run structured incremental lift studies that isolate the campaign's actual effect from what would have happened without it.
These questions matter more than the pricing model's name. Plenty of vendors will call something "performance-based" and still leave you holding all the risk.
Want a full rundown of what a real performance partner should offer? Read what to look for in a CLO platform.
Stop Paying for Impressions Nobody Can Verify
If you can tie ad spend to a verified transaction, paying for impressions is a choice to take on more risk for the same or worse outcome.
So why did we ever do it this way? Because measurement used to be hard. But it isn't hard anymore, not for anyone working with transaction-level data.
Want to see what commerce media strategies built entirely around this shift look like?
