The Questions to Ask Before Hiring an Ad Agency in 2026
August 28, 2026
The short version: the reliable signs an ad agency isn't working are behavioral, not numerical. Watch turnaround time on small changes, watch who can answer a hard question, and watch whether anything new has shipped this month. A bad month is noise. A slow month that looks exactly like the last three is the signal.
You're four or five months in, the reports still land on time, and something is off. Nobody has done anything obviously wrong. There's no blowup, no missed deadline you can point at, no invoice you'd dispute. The ads are running and yet here you are checking Ads Manager yourself on a Sunday because you want to see it with your own eyes.
That instinct is usually right and it is also the hardest thing in the world to act on, because the evidence you'd need to justify firing anyone is genuinely thin. A slow month happens. So most owners sit in the uncertainty for another quarter, then another one after that, and that's how a relationship nobody is happy with stretches out to eighteen months.
Here's what makes it decidable. The signs that matter aren't in the dashboard at all. They're in how the relationship behaves week to week, and the behavior breaks before the numbers do.
Almost every one of these stories opens the same way. Onboarding was genuinely good. Someone rebuilt the campaign structure, killed the obvious waste, shipped a batch of fresh creative and fixed a couple of tracking problems your last setup had been living with. The numbers moved and you felt good about the decision.
Then it flattened. Not a crash, just a slow settling into a rhythm where the reports keep arriving and nothing in them surprises you anymore. That flattening is so common it has its own diagnosis, and we wrote it up in why agency results drop after three months: the first sixty days spend a backlog of one-time fixes, and once that's spent you're watching what the team actually produces on its own.
The part worth sitting with is that nothing about this looks like failure from the inside. Everyone is polite, everyone is busy and the work is technically getting done. You didn't get a bad agency. You got a normal one, and normal is the problem.
Because the reporting cycle is monthly and the noise is weekly. A single week of soft ROAS could be a competitor bidding harder, a seasonal dip, an out-of-stock SKU or nothing at all. To be confident a decline is real rather than random, you need enough weeks to see through the variance, and by the time you have them you've already paid for two or three more months of the same.
Attribution stretches it further. Both major platforms document conversion windows that keep crediting sales days after the click, so the number you're looking at on the 5th isn't final yet. Meta explains its own attribution and learning behavior in the Meta Business Help Center, and Google publishes the same for conversion tracking in Google Ads Help. Neither of those is a scandal. It just means the scoreboard lags.
Behavior has no lag. How fast someone replies, who shows up on the call and whether anything new went live this week are all observable today, for free, without waiting on a report.
The useful split is between things that vary on their own and things that only happen when attention has moved somewhere else. These are real:
These are noise, and treating them as betrayal will cost you a partner who's doing fine work:
A bad month is noise. Three months that all look the same is a pattern, and a pattern is a decision.
More than any metric on the dashboard, honestly. Ask for something genuinely small: swap a headline, pause one ad set, push a new image live. Then just time it. In a healthy setup that's a same-day thing, or next-day if it needs design. When it takes a week you've learned something specific, which is that your request went into a queue behind other people's requests.
That is not laziness and it usually isn't even the fault of the person doing the work. It's arithmetic. Attention divides. A typical agency media buyer is managing ten or more accounts in the same week, so your small change is competing with nine other people's small changes, and we went through what that does to an account in how many accounts a media buyer should really manage.
The number to watch isn't the average either, it's the trend. Two days in month one and six days in month five tells you exactly where your account has drifted in the priority stack.
Try this one on your next call. Ask which ad drove the most revenue last week, and then ask why they think it worked. Someone living inside your account answers that without opening a dashboard, because they were there when it happened. They'll tell you about the hook, the audience, the thing they noticed on Thursday.
Someone who is not will do one of three things: promise to pull the numbers and follow up, pivot to a number they do have in front of them, or answer in language so general it would fit any business in any category. None of those are lies. All three tell you that your account is being reported on rather than run.
The Association of National Advertisers has spent years publishing research on how much of a media budget disappears into layers between the advertiser and the work, and you can read their material at ana.net. The version you experience day to day is much simpler. You ask a question about your own money and the answer has to be looked up.
It's less dramatic than people expect and you feel it inside the first two weeks. The person running your ads is in your Slack, they answer in hours, and they bring you things you didn't ask for.
That pattern is the single most repeated theme across our own client feedback, which is why we track it as the UM Embedded-Operator Standard: clients consistently describe their Unicorn Marketer as an embedded operator inside their company rather than an external agency contact. Tyler Finkelstein at Self-Talk Plus put it as "not traditional agency, expert works within company, Slack channels. Move faster, better results." Daniel Pike at Dermeleve described it as pairing with "a virtual in-house person" who is "highly competent, available, does much more than buying." Peter Gold at Smart SwimSuits called it an "anti-agency model" where "they align with goals, focus on results, truly act as part of team."
Notice that none of those quotes is about a metric. They're all about how the week feels, which is exactly the layer where a failing relationship shows itself first.
Sometimes it genuinely isn't, and firing them won't fix anything. If conversion tracking has been broken for months, every partner you hire will fly blind and every one of them will look mediocre. Allied Medical Training grew year-over-year revenue 70% in its first six months with their Unicorn Marketer, and the unlock was rebuilding attribution from scratch before anyone touched a campaign, which is the kind of fix that makes the previous provider look worse than they were.
The other honest cases: an offer that doesn't convert no matter who drives traffic to it, a target that was never realistic against your margins, or a direction that changes every three weeks so nothing gets long enough to work. Paid media amplifies whatever it points at. It cannot invent demand that isn't there.
So before you write the breakup email, get a second read on the account itself. If you decide to move anyway, do it in the right order and take your data with you, which we laid out in how to switch from your ad agency.
Ninety days is enough to judge the work, and it isn't enough to judge results on its own. Give a full quarter for the account restructure and the first creative cycle to run. After that, judge the behavior rather than the graph: response times, what has shipped, and whether the person on your calls knows your account without preparing for it.
No, and assuming it is will cost you a good partner. Seasonality, auction pressure, a broken checkout, an out-of-stock hero product and a change in your own offer all move the same numbers. The test is whether they spotted it before you did and told you straight. A partner who explains a bad quarter accurately is worth more than one who never has one.
Same day for a copy swap or a pause, one to two days for anything that needs new design. That's the bar in 2026 and plenty of teams clear it. A week for a headline change means your request is sitting in a queue. Track the trend across a few months rather than reacting to any single slow week.
Open the account yourself and look at the change history and the ad ship dates. Both Meta and Google log who changed what and when. Then compare that record against the last two monthly reports. If a report claims ongoing optimization and the log shows a handful of budget nudges, you have your answer in about ten minutes.
Yes, once, clearly, and in writing. Name the specific behaviors rather than the feeling: the turnaround times, the missing tests, the reporting gap. Good partners fix it inside a month and you keep a relationship you already invested in. If the response is a defense rather than a plan, you've learned what you needed and you can start looking without guilt.
The fastest way to settle it is to have someone experienced look at the account and tell you what they see. That's what the Unicorn Assessment is: a full read of your ads, funnels and creative, with custom assets built for your brand before there's any contract to sign. You'll know within an hour whether the problem is the agency, the offer or the tracking.If it turns into a match, you meet and approve the specific Unicorn Marketer yourself before any of your budget gets spent, you get a 14-day window to flag any communication issues, and you can switch people or walk away for any reason inside 60 days.
%20(1).webp)