Growth marketing is a working method rather than a channel: form a hypothesis, run a controlled test, keep what wins, kill what does not. Practised properly it is one of the most useful things an outside team can bring, because it replaces opinion with evidence. Practised as a label, it is the same channel work with a more confident vocabulary and a dashboard. The distinction is easy to test in a single conversation, and this page sets out the questions that do it plus the contract terms that keep the resulting evidence useful to you.
What separates method from vocabulary
Ask a candidate to describe their last five experiments for a client, including the hypothesis, what was measured, how long it ran, and the result. A real practice answers immediately and, crucially, includes failures, because a testing programme where everything wins is not a testing programme. Ask how they decide when a result is real rather than noise, and listen for whether sample size and duration are considered before launch or rationalised afterwards. Ask what they do with a losing test, since the value of a well-designed failure is that it removes a whole line of thinking. A team that talks about growth exclusively in terms of channels, tools and dashboards is describing execution, which may still be what you need, but you should not pay a premium for a method that is not present.
Reading a case study honestly
Growth case studies are the most inflated documents in marketing. Test them with four questions. Over what period, since a gain over a launch quarter often reflects the launch rather than the work. From what base, since large relative gains from small absolute numbers are easy. Against what counterfactual, since a business growing anyway will grow during an engagement. And what else changed, since funding rounds, product launches, seasonality and pricing changes all move the same numbers. Ask for a named reference from an engagement of comparable size and stage, and ask that reference what the agency got wrong. Under the FTC's endorsement guides, testimonials used in advertising must reflect honest experience and any material connection between the endorser and the agency has to be disclosed, which is a reasonable standard to hold a case study to as well.
Where the method actually pays
Experimentation earns its cost where volume makes results legible quickly and where the change is cheap to implement: onboarding flows, checkout, pricing presentation, lifecycle email, ad creative, landing pages. It struggles where traffic is low, where sales cycles are long, or where the tested change is expensive to build. That is why a candidate's first move on a small B2B account should usually be qualitative research and fixing obvious breakage rather than a testing calendar nobody has the volume to fill. Ask where they think testing will not work for you. An agency that names the limits of its own method is an agency you can believe about the rest, and it is the same instinct you want when comparing digital marketing and SEO services on a longer horizon, where compounding matters more than iteration speed.
Terms that keep the evidence yours
Insist that experiments are documented in a log you can read at any time: hypothesis, variants, audience, dates, result, decision. That log is the durable asset of a growth engagement and it should live in your systems rather than the agency's. Insist that analytics, tag management and testing tools are configured in accounts you own. Agree the primary metric per experiment before it runs, so success cannot be redefined afterwards. Agree how much of the budget is allocated to testing versus to scaling what already works. And agree a handover standard that includes the reasoning behind decisions, not just the settings, because the accumulated knowledge of what does not work for your audience is worth more than the current campaign configuration.
Questions people ask about growth marketing agencies
How is growth marketing different from a normal agency retainer?
In principle, the unit of work is a tested hypothesis rather than a deliverable, and the output includes knowledge as well as campaigns. In practice, many retainers use the label without the method. Ask for the experiment log; its presence or absence answers the question faster than any pitch.
Do we have enough traffic to run experiments?
Often not, and a good agency will say so. Reliable testing needs enough conversions per variant to distinguish a result from noise within a sensible timeframe. Below that threshold, qualitative research, fixing known breakage and sequential before-and-after comparisons are the honest alternatives.
What should we ask a reference?
What the agency got wrong, how they handled it, what the reporting looked like in month six, and whether the knowledge stayed with the client after the engagement ended. References confirm competence poorly and confirm working style well, which is what you actually need to know.
How should a growth engagement be priced?
Usually a retainer covering a defined experiment throughput plus implementation capacity, with media spend separate and paid directly by you. Be cautious of pricing tied to a share of ad spend, which rewards scaling budget rather than finding what works, and of pure outcome pricing on metrics the agency defines.