The problem with auditing your own account.
If you're paying an agency and starting to suspect they're phoning it in, your first instinct is probably to ask for an audit. Maybe from another agency. Maybe from an internal hire. Maybe from a freelancer.
That's often the right move eventually. It's usually the wrong move first, because audits take two to four weeks, cost money, and produce a document that needs interpreting.
Before you get there, you can learn 80% of what you'd learn from a full audit by asking four questions in a meeting. How the current agency answers tells you most of what you need to know.
The four questions.
One: What changed in the account in the last seven days?
A senior practitioner actually in the account can answer this in a paragraph. They added five new negatives to a Google Ads campaign. They paused an underperforming Meta ad set. They pushed two new creative variants into rotation. They reviewed the weekend pacing and adjusted the daily budgets.
An agency that isn't in the account weekly stumbles on this question. They pivot to strategic framing ("we're optimizing toward Q4 goals") or reach for quarterly planning language. Listen for specificity. Dates. Specific changes. A person's name who made the change.
Two: What's your worst-performing campaign right now, and why?
A practitioner engaged with the account knows exactly which campaign is underperforming. They have a theory about why. They have a plan, or a question, about what to do about it.
An agency in maintenance mode gives you a generic answer: "we're monitoring some underperformers, it's normal seasonal variance, we're optimizing broadly." That means they don't have a specific worst-performer in mind because they haven't looked recently.
Bonus signal: if they can't name a worst performer at all ("everything is tracking to plan"), either the account is in a very stable quarter or they're not looking critically. Most months have a worst performer. Pretending otherwise is a tell.
Three: What's the biggest risk to our spend right now?
A practitioner can answer this. Depending on the account: our conversion tracking is fragile because of a site migration. We have too much dependency on one creative that's starting to fatigue. Our feed pricing was off on a product SKU last week and we caught it but had to pause. The iOS 17 update changed something about event handling and we're monitoring.
An agency without active engagement gives you a non-answer: "no major risks right now, things are running smoothly." That can be true, but it's worth probing. Follow up with: "nothing at all? What would you want us to watch for?" A practitioner has something to say. An agency in autopilot doesn't.
Four: If this were your business, what's the one thing you'd do differently?
This question filters out the agencies who've stopped thinking critically.
A good answer is direct and might even be slightly uncomfortable. "Honestly, if this were mine, I'd cut the Reddit spend. It's not producing and we keep justifying it." "I'd bring on a creative partner to move faster on Meta variants; you're getting beat on velocity." "I'd push harder on the tracking audit we keep deferring because it's going to keep biting us."
A bad answer is a safe corporate platitude: "I'd probably invest more in brand-building over time to support lower-funnel efficiency." That answer can be written without ever looking at the account.
What the answers tell you.
If you get three or four good answers, your agency is probably engaged. There might be other issues (strategy, pricing, creative quality), but the basic work is being done.
If you get zero or one good answer, the agency has probably slipped into maintenance mode. They're running the account on autopilot, the way it was set up six months ago, making small tweaks every couple of weeks to justify the retainer. This isn't necessarily theft. It's often what happens to agency-client relationships that last more than a year. Work compounds less, attention drifts, the team on your account shifts, and nobody raises it because the numbers are mostly fine.
The numbers being mostly fine is not the same as the work being done. An account on autopilot produces mostly-fine numbers for months before it starts to regress visibly. By the time the numbers clearly show a problem, you've lost a lot of compounding that better work would have captured.
What to do with the answers.
If the agency answered well, say thanks and consider staying. You can tune the relationship, ask for different reporting, request more direct access to the practitioner, but the underlying engine is working.
If the answers were weak, you have three options.
One, escalate inside the agency. Some agencies respond well to "we're going to take a serious look at whether to renew at the next break point unless the engagement changes," and you get a different person assigned who actually does the work.
Two, get a real audit. If the agency has been phoning it in for months, there's probably accumulated damage. An external audit tells you where the damage is.
Three, make the move. If you've already been suspicious for a while, the four-question test just confirmed what you knew. Don't make this a drawn-out process. Find the replacement, run a clean transition, move on.
A year of mediocre paid media work costs more than people realize. Not because of any single mistake, but because of what wasn't tested, what wasn't caught, what wasn't improved. The four questions are a cheap, fast way to check whether the work is still happening. Ask them soon. Ask them again in three months.
Work with a senior practitioner.
Pacific Northwest Digital Marketing runs paid media for small and mid-sized businesses. Every engagement is run by a senior practitioner from first call through monthly reporting.
Start the conversation →