At pre-seed and seed, the marketing line is a rounding error next to the engineering one, and the first agency invoice is often the largest single marketing decision a founder has made. Choose on what the team will own, how they price it, and what exists after ninety days that did not exist before.
Most founders choose on service lists and case studies instead, which is why so many agency relationships end quietly at month four with a folder of deliverables and no working channel. At Series B that is a bad quarter. At pre-seed it is a chunk of runway you do not get back.
This guide is written for that constraint: a small budget, no marketing hire, and one founder still holding the channel personally. It covers when not to hire at all, what a few thousand dollars a month can honestly buy, and how to judge the work at day ninety. If you already know content is the channel and the question is narrower, the companion guide to picking a B2B content marketing agency goes deeper on that one decision.
When not to hire an agency
An agency multiplies something that already works. It cannot invent the thing.
If you cannot yet describe who buys, why they bought, and which channel produced the last three conversations, you do not have a marketing problem. You have a positioning problem, and handing it to an outside team means paying them to run your discovery while you lose the direct signal that discovery produces.
Three situations where waiting is the better call:
You have no repeatable channel yet. Not three experiments. One thing that produced qualified conversations more than once, even inconsistently.
Your ICP is still moving every month. Any messaging written now will be rewritten before it ships.
You cannot name three concrete deliverables you want in the first ninety days. "Own marketing" is not a brief, and an agency handed that brief will fill it with activity.
Founders often hire because the work has become unpleasant, not because it has become ready. Those feel identical from the inside and produce very different outcomes.
The three ways to buy marketing help
They are not interchangeable, and most disappointment comes from buying one while needing another.
An agency gives you a team and throughput. Best when the strategy is settled and the constraint is execution volume across several channels at once. The tradeoff is context: they are working on other accounts, and their understanding of your market will always trail yours.
A fractional operator gives you senior judgment for a few days a month. Best when you need someone to decide what to do, not to do it. The tradeoff is availability, and that nothing gets executed unless someone else executes it.
A freelancer or contractor gives you depth in one skill. Best when you know exactly what you need made. The tradeoff is that they will not tell you when you have asked for the wrong thing.
If you need judgment and execution and can only buy one, buy judgment. Wrong work executed well is more expensive than right work executed slowly.
What a small budget actually buys
The honest unit of an early-stage retainer is senior hours, not deliverables. A few thousand dollars a month buys one channel, part-time, from one or two people. It does not buy a team, and it does not buy four channels run properly at once.
Below roughly the cost of one senior day a week, nobody can do strategy and execution both. Something gets dropped, and it is almost always the strategy, because execution is what shows up in the monthly report. If that is your budget, buy the decision from a senior operator and execute it yourself, or buy execution of a decision you have already made.
Spend what you have on work that keeps producing after the invoices stop. Content and search on a domain you own keep working. Paid acquisition stops the day you pause it, which is a reasonable trade at scale and a poor one on twelve months of runway.
One more constraint worth naming out loud on the first call: tell them the budget. An agency that cannot scope to it will say so, and that is a faster answer than three weeks of proposals written for a company four rounds ahead of you.
What the pricing model tells you
Price matters less than structure, because the structure decides what the team optimizes for.
A monthly retainer buys continuity and is the only model where compounding work like search and content makes sense. It also rewards staying, so ask what happens in a month where nothing much needed doing.
Project pricing buys a defined outcome with a defined end. Clean for a website or a launch. Bad for anything that only works cumulatively, because the project ends before the compounding starts.
Performance pricing aligns incentives on paper and distorts them in practice. Whoever is paid on leads will optimize for lead count, which is not the same as pipeline. If you use it, define the metric at the far end of the funnel, not the near end.
Hourly is the most honest for exploratory work and the worst for anything ongoing, because it prices effort rather than result.
Whatever the model, ask what happens if you leave after ninety days. If the answer is that you keep nothing, you are renting attention rather than building an asset.
Questions worth asking on the first call
Most agency calls are a demo of the agency. Redirect them:
What will you own that I currently own, and what stays with me?
What are the first three deliverables, and by when?
Which channel would you start with for a company at our stage, and why that one first?
What would make you tell us we are not ready for this?
Who actually does the work, and will I speak to them or to an account manager?
Show me something that did not work, and what you changed.
The fourth and sixth are the ones that separate people. An agency that has never turned work down, or that cannot describe a failure specifically, is selling you certainty it does not have.
What a good first ninety days produces
Not a report. Artifacts you would keep even if the relationship ended.
By day thirty you should have a written description of who buys and why, drawn from actual customer conversations rather than assumption, and a positioning memo you recognize as true.
By day sixty, one channel should be running as a defined experiment, with a baseline, a hypothesis and weekly numbers. One channel done properly beats four half-run.
By day ninety, there should be a publishing or outbound cadence that continues without anyone holding it in their head, plus a library of customer proof your sales conversations can draw on.
If ninety days produces more meetings, more opinions and more content but no clearer message and nothing that repeats, the engagement is not working. That is worth saying at day ninety rather than month eight.
Red flags
Deliverables described in volume. Twelve blog posts a month is an output, not an outcome.
No named ICP in their proposal, or one that matches whatever you said on the call.
Case studies with percentages and no denominators. A 300% lift on a base of four is noise.
Reluctance to say which channel they would drop.
The person who sold you is not the person who will do the work, and nobody will say who will.
A recommendation to run paid acquisition before you can describe why anyone bought.