How to Choose a Marketing Agency

The question does not have an answer, and any page that gives you one is ranking vendors by who paid for the placement or by who is largest. Fit is situational: the firm that would transform a 40-person professional services business is the wrong firm for a pre-revenue founder and cannot pass a Fortune 500 security review. The useful question is which category matches your constraints, and then which two or three firms inside that category you want to talk to.

Pick a freelancer when the work is one discipline, well specified, and bounded in time, or when the budget is small enough that an agency's overhead would eat most of it. Pick an agency when the work spans disciplines that have to be coordinated, when continuity matters more than cost, or when nobody in-house can direct a specialist. The failure mode of hiring a freelancer is discovering that the problem was never in the discipline you hired for; the failure mode of hiring an agency is paying an account layer to coordinate work you could have specified yourself.

Give each of them the same brief and the same data access, and compare what they ask before they answer. A firm that quotes a plan without seeing your analytics is selling a template. Then ask each to name what they would not do and which type of client they turn away — vague answers there usually mean they take everything. Case studies are the weakest input available, because you cannot see the ones that were not published and you rarely learn what the client contributed.

Four things. Who exactly does the work, and are they on this call? What would make you tell me this is not a fit? How is success defined, in writing, before the work starts? And what does the first ninety days produce that I keep if we stop? The last one separates firms that build something you own from firms that build a dependency. None of these questions require you to know anything about marketing.

Ask for the baseline, the window and the denominator: what the number was before, over what period it moved, and what it is a percentage of. A percentage without a baseline is unfalsifiable, and a result quoted without a time window is usually one good month. Then ask to speak to the client, and specifically to a client whose engagement ended. Firms with real results can produce both; firms with assembled results will offer a logo wall instead.

Only if procurement requires it. An RFP is good at comparing large vendors on comparable terms and bad at everything else: it rewards firms with dedicated proposal staff, which correlates with size rather than fit, and small specialists frequently decline to respond. If you are buying judgment rather than a staffing model, a paid, scoped diagnostic from two or three firms tells you far more than ten proposals, because you see the work rather than the pitch.

Notice period, data and account ownership, and who owns what gets built. Insist that ad accounts, analytics properties, domains and CRM instances are yours and stay yours — this is where the worst exits happen. Then check whether the deliverables are defined by output or by hours, and what happens to work in progress if you leave mid-month. Minimum terms matter less than most buyers think; the exit clause matters more.

Three inside one category, not eight across all five. Deciding the category first is the step most buyers skip, and it is why comparison calls feel incomparable — an enterprise pod and a freelancer are not competing for the same job. Two weeks is usually enough once the category is settled. If a firm will not put a scope and a price in writing within that window, that is itself the answer.

When you cannot yet say what a good outcome would look like, or when your product has not sold to anyone. No category on this page fixes an offer that nobody wants, and every one of them will happily bill you while you find that out. Spend the money on getting to the first ten customers by whatever unscalable means work, then come back with data worth reading.