What to look for in a marketing agency (and what to ignore).

Choosing a marketing agency is one of the higher-stakes decisions a CMO makes. It's also the decision with the worst signal-to-noise ratio of any major procurement choice. Agency websites are a hall of mirrors. Awards are bought. References are coached. Case studies omit the years that didn't work.

Three signals that are harder to fake

First, the same senior people are on the pitch and on the work. Ask. Get it in writing. Most agency churn happens because the people who won the account aren't the people who deliver it.

Second, the agency has work they're proud of from clients they're no longer with — and they'll talk about why. A healthy agency loses clients without resentment. An unhealthy one only ever shows you the wins.

Third, the agency disagrees with you in the pitch. Politely, but visibly. Agencies that agree with everything you say in the sales process will agree with everything you say in the work, which is the opposite of what you're paying for.

What to ignore

Ignore award counts unless they're from juries you respect (Brand New, Awwwards SOTY, Cannes Lions Grand Prix — yes; pay-to-play directories — no). Ignore client logos taller than they are wide. Ignore the words 'data-driven,' 'creative,' and 'innovative' on any agency homepage; if they had a sharper word, they'd use it.

Questions worth asking in the pitch

The pitch deck is the worst place to learn about an agency. The Q&A is where the actual signal lives. A short list of questions that consistently surface useful information — and that bad agencies struggle to answer well.

• 01Which client did you lose last year, and what did you learn from it? (Refusal to name a loss is the answer.)
• 02What would you refuse to do for us, even if we asked? (No answer means they'll do whatever you ask, badly.)
• 03Show me the brief you'd write for your last winning pitch, redacted. (Reveals process artifact quality.)
• 04Who's the most junior person who'll write copy or write a brief on our account? (The answer should not be a contractor.)
• 05When was the last time you fired a client, and why? (Healthy agencies have fired clients. Unhealthy ones never have.)
• 06Walk me through a project that went sideways. What did you do? (Tests honesty and post-mortem culture.)
• 07If our CFO asked you to justify your retainer in three sentences, what would you say? (Tests whether they think about value.)
• 08What do you actively disagree with in our current marketing strategy? (Tests willingness to push back early.)

What to do with the case studies

Case studies are marketing pieces about marketing work and should be read with the same skepticism you'd apply to any other ad. The signals worth looking for: specific numbers (a 47% lift is more credible than 'significant improvement'), named clients (vague 'a Fortune 500 retailer' suggests the client didn't approve attribution), defined timeframes (a year-over-year comparison is more credible than 'over time'), and at least one acknowledgment of what didn't work in the engagement.

The signals to be wary of: rounded-up percentages that suggest someone picked the metric after the fact, results that imply causation without controls (a paid program 'driving' 200% revenue growth that coincided with the company adding two enterprise sales reps), and the absence of any contextual challenges. Real work is messy. Case studies that aren't messy are usually edited until they're aspirational, not honest.

“The agency that tells you the work was hard, the team disagreed about it, and the result was less perfect than the case study suggests is the agency that learned something. The agency with only triumphs is the agency that's still selling.”

Reference calls that actually work

Every agency will give you three references. All three will be coached. The references they don't give you are the ones worth talking to. Two strategies work to get to the real reference list. First, ask the agency for a list of every client they've worked with in the past three years, not just the ones they want you to call. Most will provide it; the ones who refuse are telling you something. Second, ask current clients for the names of the agencies they used to work with — and call those references about the current agency, not the previous one. Backchannel references through someone in your network are worth ten coached references through the agency.

The questions worth asking on a reference call are different from the questions agencies expect. 'Were you happy with the work' is useless — everyone says yes on the first pass. 'What was the hardest part of working with them' is closer. 'What would you change about the engagement' is closer still. 'If you had to do it again, would you hire them?' is the question that actually separates a 9/10 reference from a 7/10. Most coached references answer 9/10 with conviction. Real 9/10 answers come with examples.

Pricing transparency as a credibility signal

Agencies that won't quote a price range until you've signed an NDA are usually charging based on what they think they can extract, not what the work costs. Honest pricing is a credibility signal. We publish ranges on our pricing page — not because we want to compete on price, but because the agencies who refuse to publish ranges are the ones whose pricing reflects negotiation more than work.

The range itself is less important than the conversation around it. A good agency will tell you, in the first conversation, roughly what an engagement of your scope will cost, and what factors would move it up or down. A bad agency will tell you that pricing is 'highly bespoke' and run you through a three-meeting discovery before producing a number. The bespoke pricing dance is a sales tactic. The work isn't actually more bespoke at most agencies than at others.

Cultural fit, and why it matters more than you think

The conventional advice on agency selection focuses on capability — can they do the work, do they have the right experience, have they worked in your category. All true. The conventional advice underweights cultural fit, which we'd argue is the single largest predictor of whether the engagement actually works.

Cultural fit isn't about the agency being fun to work with, although that helps. It's about decision-making speed (does their cadence match yours?), risk tolerance (are they willing to push you, or do they default to safety?), and how they handle disagreement (does the senior team challenge you in week three, or do they wait for the relationship to fail before raising concerns?). Two agencies with identical capability and identical pricing can produce wildly different outcomes based on whether their working style matches yours. Test for it deliberately.

The first 90 days as a test

Treat the first 90 days of an agency engagement as a structured evaluation, not as the start of a long relationship. Define, in writing, what success in the first 90 days looks like — specific deliverables, specific decisions made, specific outcomes shipped. At day 90, review against those criteria with the agency. Most agencies will pass the 90-day review because they're still on their best behavior. The few that don't are worth catching early, before the contract renewal forces a harder conversation.

We do this with our own clients in reverse: at day 90 we ask them, in writing, whether the engagement is meeting their expectations and what they'd want different. The agencies that are confident enough to invite that conversation usually deliver. The ones who avoid it usually have something to hide. The 90-day review is a credibility signal that runs in both directions, and the agencies who treat it as a real conversation tend to be the ones worth keeping.