Start with one question
If you only have twenty minutes, ask this: “Walk me through the last campaign you killed, and what the leading indicator was that told you to kill it.” Good media buyers will answer inside 90 seconds with specifics — creative fatigue curves, CVR decay by placement, a CAC drift against a contribution-margin floor. Generic answers (“it just wasn’t performing”) tell you everything.
The 12-point scorecard
I use the same rubric for every evaluation. Four clusters, three items each:
Cluster 1 — Diagnostic depth
- Can they read a creative fatigue curve and name the threshold they act on?
- Do they discuss contribution margin, not just ROAS?
- Will they share a sample post-mortem of a failed campaign before signing?
Cluster 2 — Creative velocity
- What’s their weekly net-new creative output per brand?
- Who edits the raw footage — in-house, contractor, or the brand’s team?
- How do they distinguish a concept test from a creative iteration?
Cluster 3 — Measurement discipline
- Which attribution model do they use, and why?
- How often do they reconcile platform-reported metrics with ledger-level revenue?
- What lift test or geo holdout have they run in the last 90 days?
Cluster 4 — Operating rhythm
- Who’s the senior on the account, and how many hours a week do they actually touch it?
- Is their weekly read-out async, or a 60-minute slideshow?
- What do they do in the first two weeks before launch?
A senior buyer will earn you back their fee in a single well-timed kill. A junior one will spend your money telling you why the data isn’t clear yet.
Three red lights
These are the patterns I’ve watched sink engagements at three different companies. Each one is survivable in isolation, but if you see two inside a discovery call, walk away.
Red light #1 — The ROAS-only read-out
ROAS without a cost structure is a number-shaped object. An agency that talks about ROAS without ever mentioning contribution margin or payback period is optimising for a metric that can kill the business even while it rises. The tell: they can’t tell you the floor ROAS at which your unit economics break even.
Red light #2 — Creative is quietly outsourced
Creative velocity is the single biggest driver of paid performance in 2025. If the agency’s answer to “who makes the ads?” involves three layers of subcontracting, they’re a middleman, not a team. You want the buyer and the editor one Slack channel apart.
Red light #3 — The “give it 90 days” defence
90 days is a fair runway for a strategy to play out. It is not a fair shield for bad weekly decisions. If the first read-out is already being deferred (“too early to tell”), the buyer is either not looking at the right leading indicators or is stalling. Ask them to name the specific leading indicators they are watching in week one, and the threshold each needs to cross.
What a senior freelancer can do that a bad agency can’t
I’ve now led paid media across regional DTC, F&B, telco, and SaaS — five industries with wildly different unit economics. The single biggest pattern: the agencies that failed were optimising for effort, not outcome. They shipped more ads, more meetings, more decks. A senior freelancer who owns the P&L will ship fewer things, but each one will be closer to the dollar.
How to run the evaluation call
Block 45 minutes. Send the scorecard in advance so the agency knows what to prepare. Ask one question from each cluster. Listen for specifics, not frameworks. Save the last ten minutes to ask what they would deliberately not do in the first month — that question separates operators from sellers faster than any reference call.