How to choose between performance marketing companies

Agencies with a verified published price

How to buy performance marketing

1. Define the outcome you are paying for. Write down the conversion event, where it is recorded, and what disqualifies one. A lead that is a wrong number and a lead that books an appointment cannot both count if the fee or the bonus depends on the total.
2. Fix the fee model in writing. Percentage of spend, flat retainer, cost per acquisition and revenue share all create different incentives. Ask each bidder to quote the same programme under two models so you can see what the choice actually costs you.
3. Audit the measurement before launch. Check the conversion tags fire once, carry the right value, and exclude tests and refunds. Reconcile platform reported conversions against your own system in month one, and agree the tolerance at which a gap becomes an investigation.
4. Keep the accounts in your name. Ad accounts, analytics properties, tag containers and pixel ownership should sit with you, with the agency granted access. That single condition turns a provider change from a rebuild into a permissions update.

What separates the agencies is measurement, not channels

Almost every performance agency can run the same platforms, and platform certifications are close to universal, so they discriminate poorly between bidders. What varies enormously is measurement discipline: whether the agency reconciles platform reported conversions with the client's own records, whether it reports on incrementality or simply on last click, and whether it will show you a month where the numbers went the wrong way and explain why.

Ask for a redacted monthly report from a client the agency still serves. Read it for the losses. A report that only narrates wins is a sales document, and an agency that cannot describe a failed test in its own account has either not run one or does not tell clients about it. Both are informative.

Fee models and the incentives they create

Percentage of spend is the most common model because it is simple to administer and scales with workload, but it rewards larger budgets regardless of return, so it needs a floor, a cap or a written review trigger. Flat retainers align better with steady programmes and make shortlist comparison straightforward. Cost per acquisition and revenue share look like perfect alignment and often are, provided the definition of the acquisition is airtight and audited, because the moment a fee depends on a count, the count needs a referee.

Whatever the model, ask what happens when you cut spend for a quarter, what the notice period is, and whether creative production sits inside the fee or bills separately. Those three answers explain most of the difference between two quotes that appear to price the same service.

Reading claims about return

Treat any return on ad spend or cost per acquisition figure in a pitch as a forecast that rests on assumptions, and ask what those assumptions are: attribution window, view through inclusion, margin treatment and whether returns and refunds are netted. Two agencies quoting different numbers for the same account are usually using different windows rather than different skill.

The defensible way to buy is to agree one governing report at the start, ideally your own system of record, and to hold every review against it. Agencies that already work this way will say so unprompted, because it protects them from being blamed for a platform's optimism as much as it protects you.

Questions people actually ask

Is performance marketing the same as paid media buying?

Not quite. Media buying is the execution layer. Performance marketing describes buying that is judged and often priced on a measured outcome, which usually means the agency also touches tracking, landing pages and creative testing rather than only bids and budgets.

Should I pay on results?

Only where the result is countable, auditable and hard to game. Ecommerce purchases qualify readily. Lead volume does not, unless you define disqualification and someone independent can verify the count. Otherwise a flat fee with clear reporting is usually the cleaner arrangement.

Why do the platform numbers and my numbers disagree?

Attribution windows, view through credit, cross platform double counting and refunds all pull the two apart. Neither report is dishonest, but only one reflects money received, so name your system of record at the outset and review against it.

How long is a fair trial period?

Long enough for a full purchase cycle plus a learning period, and specified in advance. Ask each bidder to write what they expect at 30, 60 and 90 days, then judge the engagement against their own written expectation rather than a moving one.

Browse agencies by specialty

• SEO agencies with verified facts
• PPC agencies with verified facts
• Social media agencies with verified facts
• Content marketing agencies with verified facts
• Email marketing agencies with verified facts
• Performance marketing agencies with verified facts