How to choose a marketing agency, and when to keep it in-house

Most advice on this compares agencies to each other. That is the wrong first question. The first question is which model should own the work at all, because in-house, freelancer and agency fail in completely different ways, and picking the wrong shape costs more than picking the wrong supplier inside the right one. This is the version we would give a friend, including the parts that argue against hiring us.

The three models, and what each is actually good at

There are only three ways this work gets done. Everything else is a variation.

Agency is the third: a team you rent, with more coverage than a freelancer and less context than an employee. The honest trade is breadth and continuity in exchange for depth of product knowledge. A good agency closes that context gap deliberately, which is what a real onboarding is for. A bad one never notices it exists.

What each actually costs, not what it is quoted at

The quoted number is never the real number. Three costs get left out of the comparison every time.

1. Management time. An employee needs hiring, onboarding, direction and review. That is founder or manager hours, and at small company size those are the most expensive hours in the building.
2. Tooling. Rank tracking, crawlers, competitive data, reporting and testing tools are a real monthly line. Agencies amortise these across clients; in-house pays retail for a single seat.
3. The cost of being wrong slowly. The most expensive outcome is not overpaying, it is a channel run competently-but-not-well for a year. Nothing looks broken, so nothing gets fixed.

Do the arithmetic on total cost of the channel, not on the invoice. What agencies actually cost is the wrong question until you have done it. A cheaper retainer that needs eight hours a month of your attention is not cheaper.

Retainer, percentage of spend, or performance

The pricing model decides whose interests the incentives serve when the two of you disagree. That matters more than the number.

Our own bias, stated plainly: we charge flat retainers rather than a percentage, because the advice we most often need to give is stop spending on that, and no one should have to argue themselves out of money to say it.

Who actually does the work

This is the single biggest gap between what an agency sells and what it delivers, and it is structural rather than dishonest. Agencies grow by leverage: seniors win the work, juniors execute it. That model is fine when the seniors stay involved. It fails quietly when the person in the pitch is never seen again.

You are not owed a founder. You are owed the truth about who opens your account. Ask for names and roles, ask how many other accounts that person carries, and ask who you contact when something breaks on a Friday. An agency that answers those crisply has thought about it. One that answers with "the team" has not, and it is most of what makes an agency worth paying for. More on who does the work.

Twelve things that should stop you signing

Not preferences. Each of these is something we have seen do measurable damage to an account we later inherited.

1. A guarantee of rankings or a specific ROAS. Nobody controls the auction or the algorithm. A guarantee is either ignorance or a sales tactic. Why guarantees are not possible.
2. They want to own your accounts. Ad account, billing, analytics and tag manager should be yours, with access granted (why this matters). Ownership by the agency is a switching cost dressed as convenience.
3. No audit before a proposal. A price quoted before anyone has looked at the account is a price for a generic package, not for your problem. It is also why honest pricing starts with "from".
4. Reporting that is a dashboard screenshot. Numbers without interpretation are not reporting. The value is what changed, why, and what happens next, at a cadence you agreed up front.
5. The strategy is a channel list. "We will do Search, PMax, Shopping and SEO" is an inventory, not a plan. A plan says what gets funded first and what deliberately does not.
6. They cannot explain your tracking. If nobody checks whether conversions are counted correctly before optimising, every decision after that is made on numbers that may be wrong.
7. Case studies with no denominators. "400% increase" from what, over what period, at what spend and margin. Percentages without bases are decoration.
8. Contract lock-in with no exit. A minimum term is reasonable. A twelve-month deal with no break clause and no offboarding terms is designed for their cash flow, not your outcome. On contract length.
9. Pressure on the first call. Discounts that expire, slots that are nearly gone. Marketing services are not a flash sale, and urgency is a tell.
10. No named person on the account. See the section above. "The team" is not an answer.
11. They will not say what they would not do. Anyone who thinks every channel is right for you is selling capacity, not judgement.
12. No offboarding plan. Ask what you keep if you leave in six months, and know the signs it is time to go before you need them. If the honest answer is "nothing", you are renting results rather than building an asset.

What to ask on the first call

1. Who specifically will work on this, and how many accounts do they carry?
2. What would you look at first, and what would you expect to find? Tests whether they have thought about your business or their process.
3. What would make you tell me not to hire you? The most useful question on this list. There is always an honest answer.
4. If halving the budget were right, what happens to your fee?
5. What do I own and keep if this ends in six months?
6. Show me a client you did not get results for, and what you learned. Everyone has one. Only some will discuss it.

When to keep it in-house, honestly

There are cases where hiring anyone external is the wrong move, and it is worth saying so on our own site.

• Your spend is small enough that any competent retainer is a large share of budget. Below a certain point the fee eats the media, and you are better off learning the platform yourself for a few months.
• The product or buying cycle is genuinely unusual. Highly technical, regulated or long-cycle B2B often needs context that takes longer to transfer than to build internally.
• You already have a strong operator. If someone internal is good and interested, funding their tools and training usually beats adding a layer.
• Marketing is your actual product. Then the capability belongs inside the company by definition.

The mixed model is underrated and works well: keep the channel that drives most revenue in-house where the context lives, and buy outside help for the specialist and spiky work. That is often the right answer for a business past its first few hires.