What does a performance marketing agency do?
A performance marketing agency plans, runs and improves marketing against measurable actions such as qualified leads, purchases or subscriptions. Its work often includes paid media, creative testing, landing-page recommendations and measurement. You should hire one when you have a commercially credible offer and need specialist help acquiring customers at economics your business can sustain.
The word “performance” does considerable unpaid labor in agency positioning. Ask which performance, who measures it, and what happens when a better advertising number leaves you with less money.
What should be in the scope?
The agency should first establish the economics of the customer it is trying to acquire. That means understanding your margin, sales cycle, capacity and constraints before recommending a media budget.
| Workstream | What you should receive | What to clarify |
|---|---|---|
| Diagnosis and planning | Commercial objective, assumptions, channel rationale and budget allocation | Who checks margins, lead quality and sales capacity? |
| Campaign management | Account structure, targeting, bidding, exclusions and budget pacing | Which platforms and markets are included? |
| Creative | Test concepts, production requirements and learning from results | Does the fee buy finished assets or briefs for your team? |
| Conversion experience | Landing-page diagnosis and test recommendations | Who builds changes and how quickly? |
| Measurement | Event definitions, account access and reporting rules | Who fixes tracking and reconciles orders or CRM records? |
| Decision-making | Recommendations, tradeoffs and next actions | Who can change spend, offers or targeting? |
Some agencies provide all of this. Others manage media and coordinate with specialists. Both arrangements can work if the dependencies have names attached. An excluded service still needs an owner when the plan relies on it.
For a B2B company, cheaper form fills may be worthless if sales rejects most of them. For a retailer, a higher purchase count may create a fulfillment problem if the product is nearly out of stock. Those are ordinary operating facts a competent performance partner should ask about.
Performance marketing does not imply performance-based fees
Agencies can charge a fixed monthly fee, a percentage of media spend, a project fee, a success component or a combination. The name of the agency does not tell you its compensation model.
Each model creates a question. Under a percentage-of-spend agreement, what justifies the fee increase when spending rises? Under a revenue-share agreement, how are existing customers, refunds, discounts and transactions claimed by other channels treated? Under a fixed fee, what volume of creative and analysis can the team sustain?
Request separate figures for media, agency management, production, technology and pass-through costs. Google's third-party policy requires disclosure of management fees and, when sharing Google advertising cost data, separation of the amount charged by Google from the provider's fees. That is a useful baseline for a transparent commercial relationship. Google third-party policy.
ROAS can improve while the business gets worse
Return on ad spend is attributed revenue divided by advertising spend. It does not automatically subtract product cost, returns, shipping subsidies or agency fees.
Consider this hypothetical month. Ads cost $20,000 and the platform attributes $80,000 of revenue, producing 4x ROAS. After the relevant variable costs, the orders have a 30% contribution margin before marketing, leaving $24,000. Subtract media and a $5,000 agency fee and the result is negative $1,000, before additional creative costs or fixed overhead.
That calculation also assumes the attributed revenue represents additional sales. Some customers might have bought anyway. Treat the example as an economics check, not a method for proving advertising caused every order.
Agree on the business measure behind the platform target. A first-order contribution measure may suit one retailer; a longer payback window may suit another with credible repeat-purchase data and sufficient cash. A B2B company may need qualified pipeline and closed-business feedback rather than a purchase metric.
Attribution and incrementality answer different questions
Attribution assigns credit under a set of tracking rules. Incrementality asks how much additional activity the advertising caused. Google's lift studies use treatment and control groups to estimate campaign impact. Availability and study design depend on the campaign and measurement conditions. Google's explanation of lift studies.
Ask an agency to explain which conclusions its evidence supports. A dashboard can guide day-to-day optimization without proving causation. An experiment can provide stronger evidence but still have uncertainty, limited statistical power or a result that applies only to the tested period and audience.
A smaller advertiser may lack enough volume for a useful lift study. A good agency will explain that limitation, establish consistent reporting and test what can be learned responsibly. It should not manufacture scientific certainty from a week of purchases.
Judge the first 90 days by the right milestones
Use the following as an example sequence to negotiate, not a universal results timetable.
During the initial phase, establish account access, conversion definitions, commercial targets and a baseline. Check what has already been tried and why it stopped. Existing campaigns may need to stay live during the audit.
The next phase should produce a prioritized set of interventions. Each needs a hypothesis, an owner, a budget or capacity limit, and a condition for changing course. “Test more creative” is incomplete until somebody knows which customer concern the creative addresses.
At the review, examine business performance alongside delivery. What changed? What was learned? Which assumption failed? What decision follows? Account for the sales cycle; a long-cycle offer may need more time to produce revenue than a repeat retail purchase.
What to request before signing
- A relevant case study with dates, baseline, budget context and the agency's exact contribution.
- A sample report that shows a disappointing period and the decision it triggered.
- The names and roles of the people who will run the account.
- A written explanation of fee incentives and attribution rules.
- A responsibility map covering tracking, creative, landing pages and sales follow-up.
Disqualify guarantees that require the agency to control things it cannot control. Also question an agency that refuses to consider business performance beyond the platform it operates. Limited authority is a reason to define dependencies, not a reason to ignore them.
Use Agency Field Guide to find relevant providers, then apply the buyer's scorecard. Your final choice should be the team that can explain how its work is expected to improve your economics and how both sides will know when that explanation is wrong.