How to choose a performance marketing agency
Every agency calls itself performance-driven; few accept performance risk. This guide helps you evaluate performance marketing partners: the engagement models, the questions that expose weak ones, the red flags, and how to structure a pilot that tells you the truth.
The engagement models, honestly labeled
"Performance marketing" covers three genuinely different deals. Fee-based agencies charge retainers or percentages of spend — you carry all media risk; they are paid regardless of outcome. Hybrid agencies charge a reduced fee plus performance bonuses — risk is shared, incentives partially aligned. True performance partners are paid only for validated results — CPA, CPL, FTD — and carry the media risk themselves: failed tests are their cost, not yours.
None of these is universally right. Fee models make sense for brand campaigns and creative production, where outcomes resist direct attribution. Performance models make sense when you have a validated funnel and want conversions, not experiments at your expense. What never makes sense is paying fee-model prices for what is sold as performance work. The first question to any agency is therefore simple: what exactly do you get paid for, and what happens to your fee when the results do not come?
The questions that expose weak partners
Ask these before any pitch deck opens. "Which conversions do you bill for, and how are they validated?" — a performance partner answers with event definitions, windows and holds; a weak one answers with "we optimize for ROAS." "What do you need from us before launch?" — professionals list tracking, KPI definitions, caps, creative rules and approval feedback speed; amateurs need only your budget. "Show us a rejection report from an existing client" — the existence, format and honesty of that report tells you how disputes will feel. "Which traffic sources will you use?" — "proprietary mix" is acceptable for exact placements; refusal to name traffic types is not.
And the meta-question: "What happens in month one?" A serious answer is a capped pilot with defined success gates. A weak answer is a ramp plan — budgets growing on schedule, results to follow. In performance marketing, scale should follow evidence, never the calendar.
Red flags that predict trouble
Guaranteed results before seeing your funnel — nobody honest promises conversion economics they have not measured. Opacity about sources — you will never know what you are actually buying, and neither will they when it breaks. No written validation terms — approvals, rejections and holds settled "as usual," meaning settled against you later. Long mandatory contracts with volume commitments — performance partners earn renewals through results, not lock-ins. Case studies with unverifiable numbers and no client permission — real references take calls.
And the subtle one: disinterest in your unit economics. A partner who does not ask what a validated conversion is worth to you cannot price their own risk — which means they plan to learn it with your money. The best partners ask uncomfortable questions about approve rates, validation speed and funnel weaknesses before promising anything.
Structuring the pilot that tells the truth
A pilot is a measurement exercise, not a small campaign. Its design: a capped budget and volume, per GEO; written KPI terms — event, qualifiers, window, hold; tracking verified end-to-end before the first click; and validation feedback flowing fast enough for the partner to optimize during the pilot, not after. Set success gates in advance — approve rate, cost per validated action, cohort behavior — and agree what happens at each outcome: scale, iterate, or stop.
Judge the pilot on what it was for. Volume means nothing at this stage; the answers do. Did conversions validate at the expected rate? Did the partner's optimization visibly respond to your feedback? Were rejections handled against the written terms or relitigated? A pilot that ends "economics work on two of four sources, scale those" is a complete success. That is what you were buying: not conversions, but certainty.
How this maps to working with Mobadly
Our model is the performance end of that spectrum: we are paid for validated actions under written KPI terms, and we carry the media risk of the sources we run. The engagement follows the structure above — brief, written terms, capped pilot, quality-gated scaling — with source-level quality control and weekly reconciliation as part of the deal, not extras. We combine a managed buying team with our own in-app inventory, so part of your volume comes from placements we control directly.
We are also not the right answer for every situation: unproven funnels need product work before paid traffic, and some offers need breadth we would not give them. The honest starting point is the same either way — your numbers, your funnel, your terms. The offer brief takes fifteen minutes and starts exactly that conversation.