When you launch your first campaign, the platform asks how you want to pay, and the answer shapes everything. Some models bill you for eyeballs, some for clicks, some only when someone does the thing you care about. Pick wrong and you can burn a budget fast without learning much.
I have spent over a decade buying traffic, and the pricing model is not a technical footnote. It decides who carries the risk, you or the platform, and what you can even measure early on. Let me walk you through the main models the way I wish someone had explained them to me.
The models, from top of funnel to bottom
Every pricing model is really a question about what event triggers a charge. Here are the ones you will meet on US platforms like Google, Meta, TikTok and the programmatic display world:
- CPM (cost per mille): you pay per 1000 impressions. Great for awareness and reach, and it is the native currency of most auctions under the hood.
- CPC (cost per click): you pay when someone clicks. Impressions are free to you, so the platform absorbs the risk of low click rates.
- CPA (cost per action) and CPI (cost per install): you pay when a defined action happens, a purchase, a signup, or an app install. Most of the risk sits with the platform.
- CPV (cost per view): common in video, you pay when someone watches to a threshold (for example a few seconds or the whole clip).
- CPL (cost per lead): you pay per captured lead, like a form fill. Popular in insurance, home services and other US lead-gen verticals.
Notice the pattern. At the top you pay for exposure and hope it converts. At the bottom you pay only when value shows up. That single idea explains almost everything about these models.
Who carries the risk, you or the platform
Think of it as a spectrum of who is on the hook when a campaign underperforms. Under CPM, you carry all of it. If your creative gets ignored and nobody clicks, you still pay for every impression served. The platform got paid no matter what.
Under CPC, the platform shares the risk. It only earns when your ad earns a click, so it has a reason to show your ad to people likely to engage. Move to CPA, CPI or CPL and the risk shifts almost entirely onto the platform, because it eats the cost of impressions and clicks that do not convert. That is why the deeper models need a pile of conversion data before a platform will optimize toward them reliably.
None of this is charity. Platforms price in that risk, so a CPA-optimized campaign is not automatically cheaper. The platform just needs enough signal to predict conversions before it will run one for you.
How the models connect: a tiny worked example
Here is the part that clicks for most people. These models are not separate universes, they translate into each other through your metrics. The bridge is your click-through rate (CTR), the share of impressions that turn into clicks.
Say you buy on CPM at a 10 dollar CPM, so 1000 impressions cost you 10 dollars. If your CTR is 1 percent, those 1000 impressions produce 10 clicks. Ten dollars divided by 10 clicks is a 1 dollar effective cost per click (eCPC). Now double your CTR to 2 percent with better creative, and the same 10 dollars buys 20 clicks, dropping your eCPC to 50 cents. You never changed your bid. Better creative did the work.
Keep going down the funnel. If 1 in 10 of those clicks converts, your effective CPA is 10 times your eCPC. So a 50 cent eCPC at a 10 percent conversion rate is a 5 dollar effective CPA. This is why I tell beginners that CPM, CTR and conversion rate together define your real cost per outcome, no matter which model you buy on.
Where automated and value bidding sits on top
On top of these raw models, US platforms layer automated bidding. You are still charged in the underlying auction (usually CPM), but you hand the platform a goal and let its algorithm set individual bids to hit it. A few are worth knowing by name.
Maximize conversions gets as many actions as it can inside your budget. Target CPA aims for a specific cost per action, say 20 dollars a signup. Target ROAS (return on ad spend) optimizes toward a revenue multiple, for example 4 dollars back for every 1 dollar spent, which matters when your conversions have different values. These count as value bidding when they factor in how much each conversion is worth, not just whether it happened.
The catch is that automation needs food, and that food is conversion data. Feed it too little and it guesses badly. Most platforms want a steady flow of conversions per week before target CPA or target ROAS behave, so brand new accounts often start simpler and graduate later.
How a beginner should actually pick
Start with what you can measure. If you cannot yet track conversions cleanly, do not reach for CPA or target ROAS, you will just confuse the algorithm and yourself. Get tracking solid first, then let the model follow.
My rough default for someone new: use CPC or a click-focused objective while you gather data and learn which creative and audiences work. It keeps you from overpaying for impressions that go nowhere, and every dollar buys a click you can study. Once a reliable conversion signal flows back to the platform, move up to maximize conversions, and only then to target CPA or target ROAS once volume supports it.
Match the model to the goal too. Awareness for a launch, CPM is honest. Testing a landing page, CPC. Chasing signups, installs or sales with data behind you, CPA and its automated cousins. The mistake I see most is beginners picking the fanciest bidding model on day one, then wondering why it will not stabilize.
Key takeaways
- The pricing model decides who carries the risk: CPM puts it on you, CPC shares it, and CPA, CPI and CPL shift most of it to the platform.
- The models translate into each other through your CTR and conversion rate, so a strong creative can cut your effective CPC without touching your bid.
- Beginners should start on CPC while tracking and data build up, then graduate to automated target CPA or target ROAS once conversion volume supports it.