How to Find a Media Buying Partner: Four Shapes, Where They Come From, and How to Vet One

"Find a media buying partner" is one request that hides four different jobs. Before looking at anyone's portfolio, it is worth deciding which of the four you are actually hiring for — most bad partnerships are a mismatch of shape, not a failure of skill.

Four shapes, four different deals

An agency. You are buying capacity and process: multiple people, existing platform relationships, and someone accountable when a campaign stalls. You pay a management fee or a percentage of spend. Best when your budget is large enough that the fee is a small fraction of it, and when you need continuity more than you need speed.

A freelance buyer. You are buying one person's judgement, usually in one channel they know deeply. Cheaper and faster to start, and structurally fragile: they can get busy, sick, or hired elsewhere. Best for a defined push, a single channel, or a test you want run by someone who has run it fifty times.

An in-house hire. You are buying accumulation — knowledge that stays in your business. The slowest and most expensive route to first spend, and the only one that compounds. Best when media buying is a permanent function rather than a project.

A revenue-share or joint-venture partner. You are buying aligned incentive: they front the work, you share the upside. Best when your offer converts reliably and the constraint is buying capacity rather than product quality. Worst when the offer is unproven — nobody good takes revenue-share risk on something unvalidated, so the pool that says yes is adversely selected.

Picking the shape first eliminates most candidates before you have wasted anyone's time.

Where partners actually come from

The public channels are the weakest ones, and it is worth being honest about the ranking.

Referral from someone who has paid them. By a wide margin the highest hit rate, because the referrer has already absorbed the downside. This is what operator networks are for, and it is why the question worth asking in a community is never "can anyone recommend a buyer" but "who have you actually paid, and would you again."

Communities and peer groups. Not a directory, but the place referrals originate. What you are looking for is people talking about their own results in enough detail to be checkable — see media buyer communities for how to tell a useful one from a noisy one, and cross-border e-commerce networks for how operators build the relationship layer in the first place.

Competitors' former staff and vendors. People who have already run your category, often findable by working backwards from ads you respect.

Marketplaces and inbound applications. Highest volume, lowest signal. Usable if you treat the first conversation as a filter rather than an interview, and if you never skip references.

The pattern is consistent: the quality of a partner tracks how much of their track record you can independently verify, and referrals are simply the cheapest verification available.

Vetting, in the order that saves the most time

Ask what they spent, on what, and when. Not results first — spend first. A buyer who cannot immediately state monthly spend, channel and vertical for their last two engagements is either junior or vague on purpose.

Ask for the campaign that failed. How they diagnosed it, what they changed, when they stopped. This single question separates operators from presenters more reliably than any case study.

Check the vertical, not just the channel. Skill transfers across products far better than it transfers across regulatory environments and buyer psychology. Someone excellent in one category can be genuinely lost in another.

Take two references, and ask the references about the ending. How did the engagement end, and would they start it again. Endings are where the honest information lives.

Watch for the guarantee. Anyone promising a specific return is either inexperienced or selling something else. Confidence about process is a good sign; confidence about outcomes is not.

What to agree before money moves

Most partnership disputes trace back to four things nobody wrote down.

• Who owns the ad accounts, pixels and creative. Ownership decides what you keep when the relationship ends. This is the single most expensive thing to get wrong, because rebuilding attribution history is not possible.
• What data you see, and how often. Agree the reporting cadence and the metrics before the first campaign, not after the first bad week.
• How spend authority works. A ceiling, and a rule for who can raise it.
• The exit. Notice period, handover of accounts and assets, and what happens to work in progress.

A one-page agreement covering those four is worth more than a long contract that covers none of them.

A realistic first ninety days

1. Weeks 1–2: access, tracking verified end to end, and agreement on what a success looks like in numbers.
2. Weeks 3–6: a deliberately small test with enough budget to be readable. The goal is a decision, not a return.
3. Weeks 7–12: scale only what the test actually supported, and hold the reporting cadence even when results are good.

If there is no readable answer by week six, the problem is usually the measurement setup rather than the buyer — fix that before changing partners.

Frequently asked questions

Agency or freelancer for a first hire? Freelancer if you need one channel tested quickly; agency if you need coverage and continuity and can afford the fee.

How much should I budget for a test? Enough that the result is not noise in your own numbers. A test too small to read is more expensive than no test.

Is revenue-share safer because I pay for results? It shifts risk but narrows the pool sharply, and the people who accept it on an unproven offer are usually the ones you would not have chosen.

How do I check someone's claims? References who paid them, plus specifics on spend, channel, vertical and dates. Screenshots are not verification.

What is the most common mistake? Not owning the ad accounts.

The short version

Decide the shape first — agency, freelancer, hire, or revenue-share — because it determines everything else. Source through people who have actually paid the candidate. Vet on spend history, a failure story, vertical fit and references about the ending. Then write down account ownership, reporting, spend authority and exit before any money moves.

TrafficTalking exists for the first step in that list: reaching operators who can tell you who they have actually paid, and whether they would do it again.