Performance marketing has a definition problem. Ask five people and you will get five answers: paid ads, direct response, affiliate deals, "the stuff we can measure." The vagueness is expensive, because it lets teams optimise things that look like progress while the business gets no better.
This guide sets out what it actually is, the economics that should govern it, how channels fit inside those economics, and where measurement fails. It is written for people who decide budgets and have to defend them.
What is performance marketing?
Performance marketing is acquisition managed against measurable business outcomes rather than reach, impressions or awareness. The defining feature is the target, not the channel.
That distinction does real work. It means a brand campaign with a measured effect on qualified demand is performance marketing, and a search campaign optimised toward a conversion nobody has validated is not. What separates them is whether the number being optimised connects to money.
Why do the economics have to come first?
Because every other decision depends on one number: what a customer is allowed to cost. Get that wrong and no amount of channel skill rescues it — you will either overspend into unprofitable growth or underspend and lose share you could have afforded to buy.
That ceiling comes from three inputs.
Contribution margin. Revenue minus the variable cost of delivering it. Not gross margin, not revenue. This is the money actually available to pay for acquisition and everything else.
Expected customer value. What a customer contributes across their relationship with you, not just on the first purchase. In categories where value is skewed — a small group of customers producing most of the revenue — the mean is a trap, and a conservative percentile is the safer basis.
Target payback period. How long you are willing to wait to get the acquisition cost back. This is a financing decision more than a marketing one. A business funding growth from cash flow needs a much shorter payback than one with capital to deploy.
Put together:
| Input | Example value | Where it comes from |
|---|---|---|
| Expected 12-month customer value | 100 units | Cohort analysis of your own data |
| Variable cost to serve | 40 units | Finance |
| Contribution available | 60 units | Value minus variable cost |
| Target payback multiple | 1.5× | Cash position and risk appetite |
| Acquisition ceiling | 40 units | Contribution divided by multiple |
All values above are illustrative, to show the arithmetic rather than to suggest a benchmark. The point is the shape: the ceiling falls out of finance, not out of a media plan.
Which metrics actually matter?
Four, in most businesses. Everything else is diagnostic.
Blended customer acquisition cost
Total acquisition spend divided by total new customers, from your own data, ignoring which channel claims credit. Its value is precisely that it ignores attribution. When three platforms each report conversions and the sum exceeds your actual orders — which is the normal situation, not an anomaly — blended CAC is the number your bank account agrees with.
Marketing efficiency ratio
Total revenue divided by total advertising spend in the same period. Like blended CAC, it is immune to attribution disputes because both numbers come from your accounts. It is the fastest way to answer "is the whole budget working" without arguing about which campaign deserves the credit.
Payback period
How long until a customer has returned what they cost. The most under-used metric on this list. Two channels with identical acquisition costs but different payback periods are not equivalent investments: the faster one recycles cash into the next cohort and compounds.
Contribution margin after acquisition cost
The only metric here that tells you whether growth is making money. A business can hit every efficiency target and still lose money per customer if the underlying margin does not support the price it is paying.
Return on ad spend is deliberately absent. It is useful for comparing two campaigns inside one platform on the same day, and close to meaningless as a business target, because the platform chose the base it is measured on.
How should budget be allocated between channels?
Start by naming the constraint, because the two common situations call for opposite plans.
Demand-constrained. There is only so much qualified demand available at an acceptable price. Spending more pushes you into worse inventory and efficiency degrades. Here the work is expanding qualified demand — new audiences, new markets, better creative — rather than raising bids.
Capital-constrained. More profitable demand exists than you can currently fund. Here the work is proving efficiency well enough to justify more budget, and shortening payback so the same capital turns over faster.
Teams routinely apply capital-constrained tactics to a demand-constrained account, which is how efficiency collapses while everyone is working hard.
Once the constraint is named, allocation moves on evidence rather than a fixed split. Channels that hold efficiency as they scale earn more budget. Channels that degrade quickly get capped at the point where they stop paying. Where the reading is ambiguous — and with modelled conversions it often is — a holdout test settles it rather than an argument.
What role does creative play?
More than most measurement frameworks admit, and it is the part that most often gets treated as a downstream deliverable.
The mechanism is simple. Advertising platforms price attention through an auction, and the auction rewards ads people engage with. Creative that earns attention lowers the price of everything downstream. Creative that does not has to buy its way in, and the auction charges for that.
This is why media and creative belong in the same plan rather than in separate workstreams. A media result that cannot be attributed to a specific creative idea teaches you nothing reusable. Build the concept, the landing page and the buy against the same hypothesis, and a loss tells you which part of the argument failed.
Why does conversion belong in the acquisition budget?
Because a substantial share of acquisition cost is created after the click, and that share is usually owned by a different team.
Checkout and form design are the clearest case. Baymard Institute, which has run large-scale checkout usability testing for over a decade, puts the average documented cart abandonment rate at roughly 70% across the studies it aggregates — a figure that has been remarkably stable over time.1 Not all of that is fixable, and some abandonment is ordinary browsing behaviour. But the portion that is fixable is being paid for twice: once in media cost to generate the visit, and again in the margin that never arrives.
The practical consequence is that landing page and funnel work frequently returns more than an equivalent effort spent on bid management, and it compounds across every channel at once.
How do you know a channel is actually adding customers?
By turning it off, in a controlled way.
Incrementality is the share of conversions that happened because of a marketing activity and would not have happened otherwise. It is the question attribution is really trying to answer, and a controlled holdout test is the only method that answers it directly. Everything else infers it from correlation.
Holdouts cost money: for the duration of the test you are deliberately buying less. That cost is almost always smaller than the cost of scaling a channel for a year on the strength of conversions it did not cause. Retargeting is the classic case — an audience selected for already being interested, reported through a model that credits the last touch.
Where does retention fit?
It sets the ceiling that acquisition spends against, which makes it an acquisition input rather than a separate discipline.
The relationship is mechanical. Acquisition budget derives from expected customer value; expected value derives from how long customers stay and how often they return. A durable improvement in retention raises what every channel may pay, simultaneously, without any change in media skill.
The classic reference point is Frederick Reichheld's work at Bain, popularised through Harvard Business Review, which put the profit effect of a five percent improvement in retention at between 25% and 95% depending on the business.2 Treat the range as an illustration of leverage rather than a number to plan with — the effect in any specific business depends entirely on its margin structure and repeat behaviour. The direction, though, is consistent, and it is the reason retention work is usually cheaper leverage than squeezing the same percentage out of media buying.
What does good reporting look like?
One view, built on your data, leading with the agreed target and what moved it. Channel detail sits underneath for diagnosis, not on the front page — because leading with per-platform return invites exactly the argument the blended target was meant to end.
Three habits separate honest reporting from the other kind:
- Modelled numbers are labelled as modelled. Platforms increasingly estimate conversions they could not observe. Those estimates sit next to observed conversions, usually without visible distinction.
- Directional results are called directional. Not every finding is proven, and saying so is what makes the proven ones credible.
- The comparison period is stated. Almost any number can be made to look like progress by choosing the right baseline.
Where to go next
- What is MER, and when should you use it instead of ROAS?
- CAC payback period: the metric that decides how fast you can scale
- Marketing measurement in a privacy-first world
- The performance marketing glossary