Performance Marketing: Saat Hasil yang Terhitung Menyesatkan

Performance marketing is an approach in which campaigns are planned, optimized, and judged against a predefined, measurable customer action, such as a purchase or a qualified lead. Paying only for results is common, but it isn't what makes a campaign performance marketing. Here's the fact that should slow you down: in large field experiments at eBay, brand-keyword search ads produced no measurable short-term benefit, and conventional observational estimates had overstated what paid search was returning.

That gap between what gets counted and what actually changed is the real subject of this entry. The term sounds like a promise of accountability. It's closer to a contract about what counts as a result, and the people, partners, and algorithms working under that contract will deliver whatever it rewards.

Performance Marketing Defined

The clearest working definition comes from the IAB performance marketing criteria. A campaign qualifies when it's designed to drive a measurable consumer action, when the relationship between the advertising and that action can be measured, and when the results feed ongoing optimization. The IAB also notes that paying on the basis of the action is common, but it isn't required.

The Three Conditions a Campaign Must Meet

Each condition does real work. Check them one at a time.

• A defined action. The campaign starts with an outcome agreed before launch, such as a purchase, a subscription, or a lead that sales has accepted. Choosing the metric after the results arrive doesn't count.
• A measurable relationship. There has to be a credible way to connect exposure to the ad with the action. Without it, you're reporting coincidences.
• An optimization loop. Results change what happens next, whether that's bids, budgets, audiences, or creative.

Miss any one of the three and you're running digital advertising that happens to have a dashboard. That isn't a criticism. It's simply a different job.

The Broad Meaning and the Affiliate Meaning

You'll also see a narrower use of the term. The Performance Marketing Association affiliate research reflects how the phrase is used in the affiliate world, where brands compensate third-party publishers for leads or sales. In that setting, performance marketing and affiliate marketing are close to synonyms.

This entry uses the broader IAB meaning. Affiliate is one performance model among several, alongside paid search, paid social, programmatic, and retail or commerce media. A campaign bought on a cost-per-click or cost-per-thousand basis can still be performance marketing if it meets the three conditions above.

Performance, Brand, and Digital Marketing

Digital marketing is the wider environment, covering every channel and tool that runs on digital media. Performance marketing is a way of operating inside it, defined by the objective and the measurement method rather than by the channel. Brand marketing aims at different objectives, mainly recognition, memory, and preference, which take longer to show up and rarely map to a single trackable action.

The two aren't rival camps. They do compete for the same budget, and that's where the hard decisions sit. A Kellogg teaching case on Airbnb, published in April 2026, is set in late December 2020 and asks students to recommend how Airbnb should split its next-quarter marketing budget between performance marketing, aimed at immediate trackable bookings, and YouTube brand marketing, aimed at longer-term recognition and affinity. Students work through ROAS and assumption-based ROI estimates to reach a recommendation.

Worth checking before you cite it elsewhere: the public case description doesn't report the final allocation or any realized return. It illustrates the trade-off. It doesn't settle it.

How a Performance Program Runs

Most programs follow the same cycle. Define the business objective, pick the action that represents it, instrument the customer journey, run the campaign, read the results under agreed rules, then change one thing and read again. The cycle itself is simple. The choices inside it are where programs succeed or quietly drift.

Choosing the Action That Counts

The action you choose becomes the target that every later decision serves. A page view is easy to generate and far from revenue. A purchase is closer. For a B2B company, an opportunity that sales has qualified says more than a raw form submission ever will.

Why this matters: a counted lead isn't revenue, and a counted purchase isn't profit. If the action is shallow, the program can hit its target and still fail the business. Write down what qualifies, how duplicates are handled, and what happens to refunds and cancellations before the first dollar is spent.

Channels and Pricing Models

Paid search, paid social, display, programmatic, retail media, and affiliate programs can all run as performance marketing. None of them qualifies automatically. For channel-level implementation, see Arfadia's search engine marketing services and social media ads services.

Pricing is a separate decision from measurement. Common models include:

• CPM: payment per thousand impressions delivered.
• CPC: payment per click.
• CPL: payment per lead that meets an agreed definition.
• CPA: payment per action or per acquisition, so always state which one.
• Revenue share: a partner earns an agreed share of eligible revenue.

Paying per action moves some risk to whoever sells the media. It doesn't move all of it, and it can create new risks of its own.

What the Platform Actually Optimizes

Ad platforms optimize the signal you give them, not the intention behind it. A field experiment reported in a Management Science study on CPA incentives, run with a European e-retailer, found that a platform working under a cost-per-acquisition incentive targeted people who already had a high baseline probability of buying, rather than people whose behavior the ads actually changed. That's rational behavior under the contract. It just isn't the same thing as growing the advertiser's sales.

The finding comes from one setting, and it doesn't show that every platform behaves this way. It does show why automated optimization still needs a person deciding what the target should be.

Metrics That Separate Activity From Business Value

Performance metrics sit at different distances from the business. Impressions, click-through rate, and cost per click are diagnostic. Leads, purchases, and new customers are outcomes. Revenue and contribution margin are what the business actually keeps.

CPA and ROAS, and What They Leave Out

Two ratios dominate reporting. Cost per acquisition (CPA) divides cost by the number of conversions. Return on ad spend (ROAS) divides attributed conversion value by advertising cost. Both calculations are set out in the Google Ads conversion metric guidance.

Here's the thing about ROAS. It compares revenue the platform attributes to your ads with what you paid for those ads. It doesn't subtract the cost of the product, fulfillment, discounts, returns, agency fees, or the time your sales team spends closing deals.

A hypothetical example, not a benchmark: a campaign spends $10,000 and is credited with $40,000 in revenue, a ROAS of 4. If the gross margin on those orders is 25%, the campaign produced $10,000 in gross profit, which covers the ad spend and nothing else. Add returns or an agency fee and the campaign loses money, while the dashboard still shows a ROAS of 4.

A ROAS target only means something once you've worked out the margin behind it. The same logic applies to CPA. A cheap acquisition that refunds next week wasn't cheap.

Attribution Versus Incrementality

This is the part many programs skip. It's also the part that decides whether the numbers above mean anything.

What Attribution Can and Cannot Tell You

Marketing attribution models assign credit for a conversion to the touchpoints recorded before it, under whichever model you've chosen. Attribution answers a credit question: given that this sale happened, which recorded interactions get the credit? It doesn't answer the causal question of whether the sale would have happened anyway.

Measuring Lift Against a Counterfactual

Incrementality asks what changed because of the marketing, compared with a credible estimate of what would have happened without it. The standard tools are randomized holdouts, conversion lift studies, and geographic experiments. Google's Conversion Lift measurement documentation follows this logic, estimating incremental conversions against a counterfactual and reporting incremental return on ad spend.

The catch is statistical. Lewis and Rao's advertising experiments research, published in The Quarterly Journal of Economics in 2015, showed that advertising effects are often small relative to how much individual purchasing varies, so a precise estimate of returns can require very large experiments. Their evidence came from online display campaigns run with major US retailers and financial services firms. The takeaway isn't that advertising fails. It's that one small test rarely settles the question.

Plenty of marketers aren't measuring at this level yet. In the Nielsen 2025 Annual Marketing Report, 32% of surveyed marketers said they measure media spending across digital and traditional channels together. Treat that as a reported, self-described practice rather than an audited capability: it comes from a global online survey of 1,400 marketing professionals at manager level or above with annual budgets of at least $1 million, fielded from February 25 to March 6, 2025.

Where Attributed Results Overstate Impact

The best-known demonstration comes from eBay. In eBay paid search field experiments, released as NBER Working Paper 20171 in May 2014 and later published in Econometrica, Blake, Nosko, and Tadelis found that conventional observational estimates overstated paid search returns, because the people clicking were often already intending to buy. Brand-keyword ads showed no measurable short-term benefit. Nonbrand ads did influence some new and infrequent users, but frequent users, whose buying the ads didn't change, accounted for much of the spending, so average returns were negative.

Keep those findings in their box. They describe one large, established marketplace and short-term effects, and they don't prove that paid search fails for other advertisers. What they do show is how attributed results can quietly count demand that already existed.

Privacy Limits on Measurement

Measurement also depends on what users allow. Google's consent mode guide describes how tags adjust their behavior to a user's consent state, and Apple's App Tracking Transparency framework governs permission for tracking users across apps and websites owned by other companies. Both reduce the signal available to attribution. That's one more reason not to treat platform totals as a complete count.

When the Result Definition Backfires

Performance marketing breaks in predictable ways. Almost every failure traces back to the same root: the counted result and the business result drifted apart, and nobody noticed because the dashboard kept improving.

Shallow Events, Shallow Outcomes

If the target is any form submission, the system will find people who submit forms. Some will be buyers. Many won't. Lead volume can rise while revenue stays flat, and every weekly report will call it progress. Fix the definition first, then judge the channel.

Paying for Demand That Already Existed

Attribution rewards being present near the sale. Brand search and other tactics that reach people close to purchase are well placed to collect credit for buyers who were coming anyway. The eBay experiments and the Management Science study point the same direction from different angles. One shows attributed returns overstating causal returns. The other shows an optimizer under a CPA incentive favoring likely buyers over persuadable ones.

Affiliate Disclosure Is Still Your Problem

Paying partners per sale doesn't hand off responsibility for what they publish. In the US, the FTC endorsement guides FAQ says affiliate relationships and other material connections should be disclosed clearly and conspicuously. That's US guidance, not a rule for every market, so check the requirements wherever your partners actually publish.

When the Frame Doesn't Fit

Some objectives don't belong in a performance frame at all. If the goal is recognition among people who aren't shopping in your category yet, there's no near-term action that represents it, and forcing one will push spend toward whoever was already buying. The same is true when you can't measure the link between ad and action. Without that link, the second IAB condition fails. The label stops meaning anything.

Channel work is only one part of a performance program. Search and content feed the same funnel, so search engine optimization and content marketing are often measured next to paid channels, and teams that want one plan across them can look at the coordinated digital marketing services from Arfadia. The measurement cautions above apply to all of them.

Frequently Asked Questions

What is performance marketing?

Performance marketing is an approach in which campaigns are planned, optimized, and judged against a predefined, measurable customer action, such as a purchase or qualified lead. Payment tied to results is common, but it isn't required.

Is performance marketing the same as affiliate marketing?

Not always. Affiliate marketing, where brands pay third-party publishers for leads or sales, is one performance model. The broader definition also covers paid search, paid social, programmatic, and retail media campaigns that meet the same measurement conditions.

How is performance marketing different from brand marketing?

Performance marketing starts from a specific, measurable action and optimizes toward it. Brand marketing mainly targets recognition and preference over a longer horizon. They serve different objectives, and they aren't mutually exclusive.

Do advertisers only pay when a conversion happens?

No. Payment tied to results is common, especially in affiliate programs, but the IAB doesn't treat it as a requirement. A campaign bought on CPC or CPM can still be performance marketing.

Does a high ROAS mean a campaign is profitable?

Not necessarily. ROAS compares attributed revenue with ad cost only. Profit depends on margin, returns, fees, and other costs that ROAS leaves out.

How can you tell if performance marketing caused extra sales?

Attributed sales can't answer that. You need a comparison against a credible counterfactual, such as a randomized holdout, a conversion lift study, or a geographic experiment, and enough scale to make the estimate precise.

Related Terms

• Cost per acquisition (CPA)
• Return on investment (ROI)
• How marketing attribution works
• Conversion rate optimization (CRO)

Sources

• IAB performance marketing definition, Interactive Advertising Bureau, "Defining the Next Generation of Performance Marketing."
• PMA affiliate marketing industry study, Performance Marketing Association, 2025.
• Google Ads conversion metric guidance, Google Ads Help.
• Google Ads Conversion Lift documentation, Google Ads Help.
• Google consent mode implementation guide, Google for Developers.
• Apple App Tracking Transparency documentation, Apple Developer.
• NBER Working Paper 20171 on eBay, Blake, Nosko, and Tadelis, 2014; published in Econometrica, 2015.
• Kellogg Airbnb brand versus performance case, Kellogg School of Management, 2026.
• Management Science CPA incentives experiment, INFORMS.
• Lewis and Rao 2015 study, The Quarterly Journal of Economics, 2015.
• FTC endorsement guides FAQ, Federal Trade Commission, "FTC's Endorsement Guides: What People Are Asking."
• Nielsen 2025 Annual Marketing Report release, Nielsen, May 2025.