How to Hire a Marketing Agency Without Getting Burned Again

You hired an agency once. Maybe twice. The first ninety days looked fine: a slick dashboard, a standing Thursday call, a monthly report thick with impressions and "engagement rate." Then you asked the only question that pays your bills (how many customers did this bring in, and what did each one cost) and the room went quiet. By month six you were paying $4,000 a month to feel informed and no closer to profitable. On the way out you learned the ad account, the pixel, and the Google Analytics property were all registered under the agency's name. You left with a logout screen.

So now you are evaluating agencies again, ours included, and you should be suspicious. Good. Suspicion is a hiring skill. Below is the checklist we hand people who are interviewing us, with our own answers next to every question, because the fastest way to prove we are not the last agency is to pass the test we wrote.

Who owns the keys

Four assets decide whether you are a client or a hostage: the ad accounts, the analytics, the domain, and the source files. Ask, in writing, who the registered owner of each one is on the day you sign and on the day you leave.

The trap is subtle. Plenty of agencies will happily "run" your Meta and Google accounts while quietly owning the Business Manager that contains them. When you cancel, they revoke access and keep the pixel history, the audiences you paid to build, and the conversion data. Same story with a site built on their private theme or a page builder locked to their seat. You rented a house and they kept the deed.

Our answer: you own all four, from day one, in your name. We operate inside your Meta Business Manager and your Google Ads account with delegated access. The domain stays on your registrar. Code lands in your GitHub, your Vercel, your CMS. If you fire us on a Tuesday, nothing turns off on Wednesday. That is the whole premise of owning your acquisition engine.

What happens on cancellation day

Ask it plainly: "Walk me through the first 48 hours after I cancel." A good agency has a boring, specific answer. A bad one has a speech about partnership.

You are listening for three things. One, do the assets stay live and in your control, or does something break. Two, do you get the source files, the raw ad data, and the documentation, or a zip file that never quite arrives. Three, is there a clause that holds your account hostage over a final invoice. That last one is common and quietly brutal.

Our answer: there is no offboarding drama because there is nothing to hand back. It already lives in your accounts. We remove our access, export a plain-language runbook of every automation and where it lives, and stay reachable for two weeks for questions. Month-to-month after the first term. If you want the full case against lock-in dressed up as loyalty, it is the retainer trap.

If the honest answer to "what breaks when I leave" is "nothing," you hired a partner. If it is "quite a lot," you hired a landlord.

Reporting that ties spend to revenue

Here is where most of the money leaks, and it stays invisible until someone rebuilds the plumbing.

Impressions, reach, clicks, and "engagement" are not results. They are the inputs to results. The only report that matters starts at ad spend and ends at booked, paid revenue, with cost per acquired customer and return on ad spend in between. If an agency cannot draw that line, they are not measuring your business; they are measuring their own activity.

You might be thinking this is fine for a business big enough to have clean data, but yours is too small or too local to bother. Read the next part carefully, because it is a single med spa in Westchester, and it is exactly your scale.

Skin and Self came to us with the symptom you would expect: months of ad spend and no trustworthy number for what it returned. We pulled the hood and the tracking was dark. The browser pixel was never actually initializing on load, so a chunk of conversions reached Meta through one leg only. Server-side bookings carried no session identity, so the platform could not connect a Tuesday facial to the ad that sold it. Free consultations were firing a value of zero into the reporting, quietly poisoning the return math. The dashboard looked full. It was measuring almost nothing.

We rebuilt attribution from the booking backward: every completed booking and payment tied to the visitor session that produced it, values corrected, walk-ins and datacenter bot traffic filtered out so the numbers described real Westchester humans instead of single-pageview noise from Singapore. When the plumbing was honest, the real picture surfaced: $1.3M in attributed revenue at a 6.7x return on ad spend. Nothing about the spending changed that week. The measurement did. You cannot cut what is losing or feed what is winning until the report tells the truth. The full write-up is in the Skin and Self case study.

If you want to pressure-test your own reporting against this standard, book a call and bring your last agency report. Ten minutes usually tells the story.

Our answer on reporting: every account gets a spend-to-revenue view, refreshed on a schedule you control, that names cost per customer and return on ad spend. When the source data is broken, fixing the measurement is the first invoice, not an upsell, because reporting on bad data is theater. We wrote the longer version as attribution without the lies.

Scope words that mean nothing

Read the proposal with a red pen and circle every verb that cannot be verified. "Manage your social." "Optimize your campaigns." "Handle SEO." None of these describe a deliverable. They describe a mood. "Manage" can mean forty hours or four. "Optimize" can mean a genuine restructure or nudging one budget slider on the 28th so the monthly report has a screenshot.

Make them convert mood into nouns and numbers. How many ads, refreshed how often. Which specific events are we tracking, and where do they land. What is the reporting cadence, and what fields are in it. What does month one produce that month zero did not. If the scope cannot survive that translation, it was written to be unfalsifiable on purpose.

Our answer: our scopes read like a build spec. Named deliverables, named tools, named owners, a cadence, and a definition of done. You can hold the document up against the month and check it off. If you want the shape of that, our pricing and recent work both spell it out.

The questions that make bad agencies flinch

Save these for the end of the call and watch the body language:

• "Can I see this reported as revenue and cost per customer, not impressions?"
• "Everything gets built in accounts I own, correct?"
• "What is your month-to-month option?"
• "Who specifically does the work, and can I meet them?"
• "Show me one account where the numbers went the wrong way and what you did about it."

A confident operator answers all five without breaking eye contact. The last one is the tell. Everyone has a campaign that underperformed. An agency that claims otherwise is either new or lying, and an agency that can narrate the recovery is one that has actually run money with something at stake.

You were burned once because the incentives were hidden inside words like "manage" and structures like your assets sitting in someone else's name. None of that was bad luck; it was a business model. The fix is not blind trust. It is a checklist you can hold up to any agency, us included, and a set of answers that either survives contact or does not.

Ours is built to survive it. Book a call, bring the checklist, and make us earn the yes.