Marketing companies for startups: what to buy at each stage, and what to keep in house

Which marketing company a startup should hire depends entirely on stage, and the most common expensive mistake is buying execution before the message is settled. Before product market fit, keep positioning, pricing and customer conversations in house and buy only narrow technical help: analytics setup, a website build, design. After early traction, a fractional marketing leader is usually better value than a full service agency, because the problem is still what to do rather than how to do more of it. Once a channel has been proven to work, that is the moment a specialist agency earns its fee, because it can scale something known rather than guess at something unknown. Full service retainers suit startups with a repeatable motion and a budget large enough that the agency assigns senior people to the account.

This page covers choosing and structuring outside marketing help. For evaluating agencies on evidence and fee models, see a page on judging agency proposals.

Four kinds of help, and the job each one does

Option	Best for	What you get	What it needs from you
Fractional marketing leader	No senior marketer in house, strategy undecided	Positioning, channel choice, hiring plan, oversight of contractors	Founder time weekly, and authority to make decisions
Specialist agency	One channel already showing signal	Depth in paid, SEO, lifecycle or content, and a team that has done it before	A clear target, clean data, and fast approvals
Full service agency	Several channels running, repeatable motion	Coordinated execution under one contract	A real budget and someone in house to own the brief
Individual contractors	Defined deliverables with a deadline	Low cost, fast start, direct relationship	Your management attention, as nobody coordinates them

What to buy at each stage

• Pre product market fit. Founders should be doing the selling and the customer interviews. Nobody outside can learn your market faster than you can at this point, and the words customers use are the raw material for everything later. Outsource the website, design, brand assets and analytics. Keep message and pricing.
• Early traction, a handful of paying customers. Bring in a fractional leader for one or two days a week to pick the two channels worth testing and set up measurement properly. Hire a contractor to execute each test. The deliverable here is evidence about channels, not volume.
• One channel working. Hire a specialist agency in that channel with a clear target and a trial project. Keep the second channel as an in house experiment so you are never dependent on one route.
• Scaling with a known motion. This is where a full service agency or your first senior in house hire makes sense. If you can only fund one, the in house hire usually compounds further, because the learning stays.

Across every stage, keep three things in house permanently: the customer relationship, the data, and the knowledge of why things worked. Outsource the work, never the learning.

How to run the selection in two weeks

• Day one, write one page. The commercial goal in a number, the audience, what you have tried, the budget, the decision date, and who decides. Send the same page to everyone.
• Days two to four, build a long list of six. Ask founders in your network who they actually use, look at who works with companies at your stage, and read written client feedback rather than ranking positions.
• Days five to eight, one call each. Ask what they would stop first, what they need from you, and what they would report in month one. Cut to three.
• Days nine to eleven, references. Two current clients and one former client per finalist. Ask the former client what changed and what they would check next time.
• Days twelve to fourteen, buy a trial. A paid two to four week project with a deliverable, a deadline and a success measure. You learn more from that than from any proposal.

Contract terms that protect a startup

• Thirty day notice after any initial term. Long lock ins suit the agency and punish a company whose plan changes quarterly.
• You own everything. Ad accounts, analytics property, tag manager, domain, CMS, design source files, ad creative, audience lists and documentation. Write it in, and check the accounts are in your billing from day one.
• Named people and hours. Who works on the account, how many hours, and who attends the weekly call in month four.
• Scope in plain language with the deliverables and the cadence, plus what triggers an extra fee.
• Reporting defined at signature, including the metrics and who presents them.
• Handover clause. On exit you receive credentials, documentation and work in progress within a set number of days.

Warning signs, and promises to refuse

Be careful with anyone who guarantees a ranking, a lead volume or a customer acquisition cost before seeing your product and your data. Google says plainly that nobody can guarantee a number one ranking, and an acquisition cost in a live auction is no more controllable. Other signals worth acting on: a pitch deck of logos with no numbers and no named contacts, a refusal to let you own the ad accounts, a proposal that starts with deliverables instead of a diagnosis, a team you meet once, and an offer to work for equity in place of a plan you believe in anyway.

Budget, worked from the economics

Set the budget from your unit economics rather than a percentage. Take gross margin per customer, decide what you will pay to acquire one and over what payback period, multiply by your target new customers, and that is the ceiling across fees and media. Then split it. A reasonable early split is the majority on media and testing and the remainder on fees, because at this stage you are buying evidence. As a channel proves out, the fee share rises because the expertise is doing more work.

Should an early startup hire an agency or a fractional CMO?

A fractional leader, almost always, when the open question is what to do. Agencies are built to execute a known plan well. Bring the agency in once the plan exists and one channel has shown signal.

Do agencies that specialise in startups do better work?

What matters is whether they have worked at your stage, with your budget, on your motion. An agency used to large brand budgets will apply a process your company cannot absorb. Ask for two clients at your stage and speak to both.

How long should the first contract be?

Start with a paid trial of two to four weeks on a defined deliverable, then a three month term with thirty day notice. Judge paid media inside six weeks and content or SEO over a quarter, with the milestones written down at signature.

What should never be outsourced?

Positioning, pricing, the customer conversations, and ownership of your data and accounts. Those four decide whether everything else works, and they are the assets that stay with you when a contract ends.

Is paying an agency in equity a good idea?

It aligns incentives on paper and it dilutes you for work you could buy with cash, while removing the commercial pressure that keeps an agency sharp. If the work is worth doing, pay for it. Keep equity for people joining the company.

What the agency's brief leaves out: the assistant's shortlist

You now know what to buy at each stage. Before any agency pitch, ask an assistant the question your future customers will ask, and write down which startups and tools it names. Those names are the competitors your customer is actually shown.

It makes a sharper brief than a feature list would. AI Knows Us runs the question across several assistants and returns the names, so the agency starts from the same picture you hold.