CPA Marketing: Why the Model Rewards the Wrong Behaviour

CPA marketing is a performance-based model where advertisers pay only when a specific action is completed, whether that’s a sale, a lead, a sign-up, or a download. On paper, it’s the most accountable structure in paid media. In practice, it’s one of the most misunderstood.

The appeal is obvious: you only pay for results. But what counts as a result, who gets credit for it, and whether that result was going to happen anyway are questions the model quietly sidesteps. That’s where most CPA strategies start to unravel.

I spent the better part of my early career obsessed with lower-funnel performance metrics. Cost per acquisition was the number that mattered. Everything else felt soft by comparison. It took running larger budgets across more complex businesses before I started asking a question that should have come earlier: how much of what we’re paying to acquire was going to convert anyway?

What Is CPA Marketing, and How Does It Actually Work?

CPA stands for cost per action, sometimes called cost per acquisition. It’s a pricing and measurement model where an advertiser pays a fixed or variable fee only when a defined action is completed by a user. That action could be a purchase, a form fill, a free trial sign-up, an app install, or any other trackable conversion event.

The model operates across several different channel types. Paid search and paid social platforms allow you to bid on a target CPA, using their machine learning to optimise delivery toward users most likely to convert. Affiliate networks connect advertisers with publishers who promote their products in exchange for a commission on completed actions. Programmatic display can be bought on a CPA basis through certain demand-side platforms. Influencer and creator partnerships are increasingly structured with performance components, including CPA arrangements.

In each case, the mechanic is broadly the same: define the action, set the price you’re willing to pay for it, and let the system or partner optimise toward that outcome. The advertiser carries less upfront risk because spend is tied directly to results. The publisher or platform carries more, because they only earn when the action happens.

That risk transfer is what makes CPA attractive. It’s also what creates the incentive problems.

Why CPA Looks More Accountable Than It Is

The promise of CPA is clean accountability. You pay a defined amount, you get a defined result. The commercial logic is neat. The problem is that marketing accountability isn’t really about whether an action happened. It’s about whether your marketing caused it.

Attribution is where this gets complicated. Most CPA models, particularly in affiliate marketing, assign credit to the last touchpoint before conversion. If someone spent three weeks reading reviews, watching YouTube comparisons, clicking on a retargeting ad, and then finally converted through an affiliate coupon link, the affiliate takes the commission. The rest of the funnel gets nothing. The advertiser pays for an action that was, in all likelihood, already decided.

I’ve seen this pattern across multiple client programmes. One retail client I worked with had an affiliate programme that looked excellent on paper. CPA was well within target. Volumes were steady. But when we ran incrementality testing, stripping out the affiliate channel for a controlled cohort, conversion rates barely moved. The affiliates were capturing the tail end of a experience the brand had already won, and being paid handsomely for it.

This isn’t unique to affiliate. Branded paid search has the same structural issue. Brand terms convert at low CPA because the person searching for your brand name was already planning to buy. You’re paying to intercept intent you created through other means. The CPA looks efficient because the denominator is a nearly-certain converter, not because the channel created the demand.

For a sharper look at how demand creation and market penetration interact, the Semrush breakdown of market penetration strategy is worth reading alongside any CPA planning exercise.

The Optimisation Trap: Efficient Channels That Don’t Actually Grow You

One of the persistent tensions in growth strategy is the gap between efficiency and effectiveness. CPA marketing, when managed well, is very good at efficiency. It is often poor at effectiveness in the broader sense of building a business rather than just capturing conversions from people already in market.

When I was growing an agency from around 20 people to over 100, the clients who scaled fastest were not the ones who optimised hardest on CPA. They were the ones who used CPA as a floor, a minimum standard of commercial discipline, while continuing to invest in channels that reached people who weren’t yet thinking about them. The clients who treated CPA as the ceiling, cutting anything that didn’t convert immediately, tended to plateau.

This is the optimisation trap. You get better and better at converting the people who were already going to convert, while the pool of new potential customers sits untouched. The CPA dashboard looks healthy. The business stops growing.

There’s a useful way to think about this. Imagine a clothes shop. A customer who walks in, browses, and tries something on is far more likely to buy than someone who’s never been inside. The shop assistant who helps them is doing important work. But the shop still needs people to walk through the door in the first place. CPA marketing is excellent at helping the shop assistant. It does almost nothing to bring new people through the door.

This is a core theme in go-to-market and growth strategy: the channels and models that look most efficient in the short term are frequently the ones doing the least work to build long-term commercial value.

How CPA Affiliate Marketing Works in Practice

Affiliate CPA is the oldest and most established version of the model. An advertiser joins an affiliate network, sets commission rates for defined actions, and publishers sign up to promote their products. Publishers earn when their traffic converts. The advertiser pays only on conversion.

In categories like financial services, travel, retail, and software, affiliate programmes can be substantial revenue channels. The ecosystem includes content publishers, comparison sites, cashback platforms, coupon aggregators, email marketers, and increasingly, social creators.

The model works well when publishers are genuinely influencing decisions. A well-written comparison article that helps someone choose between two products is doing real marketing work. A cashback site that catches someone at the checkout stage, after they’ve already decided, is doing almost none. Both get paid the same CPA.

Managing affiliate programmes well requires more than setting commission rates and checking the dashboard. It requires understanding the affiliate mix, which publisher types are genuinely driving new customers versus which are intercepting existing intent. It requires regular auditing of attribution paths. It requires willingness to restructure commission tiers to reward incremental value rather than last-click volume.

Most programmes don’t do this work consistently. They optimise for volume at target CPA and treat the affiliate channel as a relatively passive revenue source. That’s a reasonable short-term position, but it creates structural dependency on a channel that may be delivering less incremental value than the numbers suggest.

Creator-led affiliate is changing some of this. When a creator builds genuine audience trust and recommends a product with authentic context, the affiliate link is the natural endpoint of real influence. Later’s work on creator-driven go-to-market campaigns highlights how creator partnerships, when structured well, can operate as both brand and performance channels simultaneously.

CPA in Paid Media: Smart Bidding and Its Limits

In paid search and paid social, CPA bidding is now the default recommendation from most platforms. Google’s target CPA and target ROAS strategies, Meta’s cost cap and bid cap options, and similar tools on other platforms all use machine learning to optimise delivery toward users most likely to complete the defined action at or below the target cost.

These tools are genuinely powerful. They process signals at a scale no human bidder can match. When the training data is good, the conversion window is appropriate, and the target is set realistically, smart bidding can significantly improve efficiency compared to manual approaches.

The limits are worth understanding clearly. Smart bidding optimises for the signal you give it. If you optimise for purchases, it finds purchasers. If those purchasers have high return rates, low lifetime value, or are concentrated in a narrow segment of your potential market, the algorithm has no way of knowing unless you tell it. The system is excellent at finding people who look like your past converters. It is indifferent to whether those converters represent the customers you actually want.

I’ve worked with businesses where optimising aggressively on CPA drove acquisition costs down while simultaneously narrowing the customer base. The algorithm found the easiest converters, who happened to be a specific demographic that wasn’t the brand’s target. Volume held up. Quality declined. It took months to identify the pattern because the CPA dashboard looked fine throughout.

The fix is usually to feed better signals. Optimising toward qualified leads rather than raw leads. Using customer lifetime value data to weight conversion events. Importing offline conversion data to give the algorithm visibility of what happens after the click. These approaches improve the quality of what smart bidding optimises toward, not just the volume.

Setting a CPA Target That Actually Makes Commercial Sense

One of the most common errors in CPA marketing is setting targets based on what the channel can achieve rather than what the business needs. The right CPA target comes from the economics of the business, not from benchmark reports or competitor estimates.

The starting point is customer lifetime value. If you know what a customer is worth over their relationship with your business, you can work backward to a maximum acquisition cost that preserves acceptable margins. That maximum cost is your ceiling. Your target CPA should sit below it with enough headroom to account for variability in conversion quality and customer behaviour over time.

The formula is straightforward. If a customer generates £400 in gross profit over their lifetime and you want to maintain a 4:1 LTV to CAC ratio, your maximum acquisition cost is £100. Your target CPA should be set below that, accounting for the fact that not every acquired customer will reach average LTV.

Where this breaks down is in businesses that don’t have reliable LTV data, or where LTV varies significantly by acquisition source. A customer acquired through organic search may behave very differently from one acquired through a cashback affiliate. If you’re blending these into a single CPA target, you’re making decisions based on an average that may not represent any individual channel’s actual commercial contribution.

The more sophisticated approach is to set channel-specific CPA targets based on the observed downstream behaviour of customers from each source. This requires linking acquisition data to CRM and revenue data, which is more work, but it’s the only way to know whether a channel’s CPA is genuinely healthy or just superficially efficient.

BCG’s frameworks on scaling commercial operations, including their work on scaling agile commercial teams, point to the same principle: the metrics you optimise toward need to be connected to genuine business outcomes, not just the nearest measurable proxy.

Incrementality: The Test CPA Programmes Rarely Run

Incrementality testing is the practice of measuring what would have happened without a specific marketing activity. It’s the closest thing the industry has to a genuine test of whether a channel is creating value or just claiming credit for value that would have existed anyway.

For CPA marketing specifically, incrementality testing is the single most important analytical exercise you can run. It answers the question that CPA dashboards cannot: of all the conversions we paid for, how many would have happened without us?

The mechanics vary. Geo-based holdout tests split markets and measure conversion rates with and without a channel active. User-level holdout tests use platform tools to exclude a random sample of the target audience and compare conversion rates. Both approaches require careful design and sufficient volume to produce statistically meaningful results.

When I’ve run these tests across client programmes, the results have consistently been more sobering than the CPA dashboard suggested. Channels that looked efficient frequently showed incrementality rates well below 100%, meaning a significant proportion of the conversions they claimed would have happened regardless. That’s not necessarily a reason to cut the channel, but it changes the economics materially.

If a channel claims 1,000 conversions at a CPA of £50, but incrementality testing shows only 40% of those conversions were genuinely driven by the channel, the true incremental CPA is £125. That’s a very different commercial picture from the one on the dashboard. Tools that help capture user behaviour signals, like the feedback and growth loop frameworks from Hotjar, can support this kind of deeper diagnostic work by surfacing where users are actually making decisions.

Where CPA Marketing Genuinely Works Well

None of this is an argument against CPA marketing. It’s an argument for using it with clear eyes about what it can and can’t do.

CPA models work well in categories with high purchase intent, where users are actively searching for a solution and the marketing job is primarily to be present and persuasive at the moment of decision. Financial services, insurance, software subscriptions, and high-consideration retail all fit this profile to varying degrees.

They work well when the affiliate or partner mix genuinely includes publishers who influence decisions earlier in the experience, not just cashback and coupon sites that intercept at the end. Content-led affiliates, review platforms, and comparison tools that users consult during the consideration phase are doing real commercial work.

They work well when the conversion event is closely tied to actual business value. Optimising for purchases is better than optimising for clicks. Optimising for qualified purchases, where quality signals are fed back into the bidding system, is better still. The tighter the relationship between the CPA event and real commercial outcomes, the more reliable the model becomes as a decision-making tool.

And they work well as part of a broader channel mix that includes investment in reach and awareness. CPA channels harvest the demand that brand and content channels create. When the full funnel is working, the CPA numbers look good because the upstream activity is generating qualified intent. Strip out the upstream investment and the CPA efficiency gradually deteriorates as the pool of warm prospects shrinks.

This is why growth strategy, in the fullest sense, has to account for the whole funnel. The Crazy Egg overview of growth strategy principles makes the point that sustainable growth requires both acquisition efficiency and demand generation, not one at the expense of the other.

Building a CPA Programme That Rewards Incremental Value

If you’re running or rebuilding a CPA programme, the structural decisions matter as much as the tactical ones. A few principles that hold up across different business types and channel mixes:

Define the action carefully. The conversion event you optimise toward shapes everything downstream. A form fill is not a sale. A trial sign-up is not a paying customer. If your CPA target is built around a proxy event, make sure the relationship between that proxy and actual revenue is stable and well understood. When it drifts, your economics drift with it.

Segment your affiliate mix by publisher type. Cashback, coupon, and loyalty sites operate very differently from content publishers and comparison platforms. Treating them identically in your commission structure means you’re almost certainly overpaying for last-click interception and underpaying for genuine influence. Review your attribution paths regularly and restructure commission tiers to reflect where value is actually being created.

Run incrementality tests before making significant budget decisions. If you’re planning to scale a CPA channel, or considering cutting one, incrementality data should be part of the decision. It’s not infallible, and it requires volume to be reliable, but it’s a far better basis for commercial decisions than last-click attribution alone.

Connect acquisition data to downstream revenue. If your CPA programme is disconnected from your CRM and revenue data, you’re flying partially blind. Customers acquired through different channels, at different CPAs, often have materially different lifetime values. Without that connection, you can’t know whether your CPA targets are genuinely healthy or just superficially efficient.

Treat CPA as a constraint, not a strategy. A target CPA is a commercial guardrail. It tells you the maximum you’re prepared to pay for an action. It doesn’t tell you how to grow, which audiences to reach, what to say, or how to build lasting commercial advantage. Businesses that confuse the constraint for the strategy tend to find themselves very efficient at a shrinking opportunity.

BCG’s research on go-to-market strategy and commercial planning makes a consistent point across industries: the businesses that scale well are the ones that maintain commercial discipline while continuing to invest in the activities that build future demand, not just capture current intent.

There’s a broader point here that connects CPA to the wider discipline of growth strategy. If you want to think through how performance marketing fits into a complete go-to-market approach, the Go-To-Market and Growth Strategy hub covers the full picture, from positioning and channel strategy through to measurement and scaling.

The Honest Assessment: CPA Is a Tool, Not a Strategy

I’ve judged marketing effectiveness awards. The campaigns that win aren’t the ones with the lowest CPA. They’re the ones that moved business outcomes in ways that mattered, reached new audiences, changed perceptions, created demand that wouldn’t otherwise have existed. CPA marketing rarely features in those cases, not because it isn’t valuable, but because it’s a harvesting mechanism, not a growth engine.

That’s not a criticism. Harvesting is important. If you’ve created demand through brand investment, content, PR, and product experience, you need efficient mechanisms to convert that demand into revenue. CPA channels, when well-managed, do that job effectively. The error is in treating the harvesting mechanism as the whole of marketing strategy.

The businesses I’ve seen stall on CPA-led strategies share a common pattern. They optimise their way to efficiency, cut the channels that don’t convert immediately, and find themselves with a very tight, very efficient funnel that gradually runs out of new entrants. The CPA looks great right up until the volume starts to drop, and by then the brand investment needed to refill the top of the funnel has been defunded for long enough that recovery takes years, not months.

CPA marketing is worth understanding thoroughly, managing rigorously, and using honestly. It is not worth mistaking for a complete commercial strategy. The model rewards the measurable, the attributable, and the immediate. Growth, in the fullest sense, requires more than that.

About the Author

Keith Lacy is a marketing strategist and former agency CEO with 20+ years of experience across agency leadership, performance marketing, and commercial strategy. He writes The Marketing Juice to cut through the noise and share what works.

Frequently Asked Questions