How to evaluate a Facebook ads agency before you sign
Evaluate a Facebook ads agency on five things: which number they will report and how it is calculated, who works on your account and how many others they hold, what month one looks like, the notice period, and whether they will show you an account they inherited rather than built.
Which number will you report, and how is it calculated?
Ask for the definition, not the label. ROAS reported inside Meta counts the conversions the platform attributes to itself, in its own window. Blended ROAS divides all revenue by all spend across every channel. New customer ROAS strips returning buyers out. Three different numbers, three different stories, and a firm can be honest while reporting any of them.
A good answer names the metric, states the attribution window, says whether the source of truth is Shopify or the ad platform, and what happens when the two disagree. It also says what they would report in a bad month. An evasive answer stays at the level of “we look at everything holistically”, or hands you a dashboard without saying which field drives the headline figure.
Who works on my account, and how many others do they hold?
The person in the pitch is often not the person in the account. That is normal and not something to hold against anyone, but you want to know before signing, not in week three.
Ask for a name, a title, and how many other accounts that person carries. A good answer gives both without hesitating, and is straight about holidays and handovers. An evasive answer describes a pod, or a team of specialists, with no individual attached to your account.
The number itself is a tradeoff rather than a score. Fewer accounts per operator costs the firm capacity and buys you attention. More accounts per operator lowers your fee. Both models are legitimate, and how Plaid Testing runs paid media sits on the senior operator side of that line, which is a cost as well as a benefit.
What happens in month one, and how do I leave?
Month one should be diagnosis, not a rebuild. A good answer describes reading the account before changing it, checking that tracking is intact, and naming what they will deliberately not touch until they have data. An evasive answer promises a full restructure in week one, which resets learning and erases the baseline you would have judged them against.
Then ask to see an account they inherited rather than one they built from scratch. Accounts built from nothing show taste. Inherited accounts show judgment, because inheriting is what they will be doing with yours.
Finally, the notice period. Thirty days is common, ninety exists, minimum terms are normal. None is wrong on its own. What they tell you is how much the firm expects to earn your renewal.
When an agency is the right answer, and when it is not
An agency is the right answer when the constraint is execution capacity. If you know what you are selling, to whom and at what margin, and the gap is that nobody has time to build campaigns, brief creative and keep the account tidy, an agency is built precisely for that, and a good one delivers at a scale one internal hire cannot match.
It is the wrong answer when the constraint sits above the account: the offer, contribution margin per customer, the total budget, or whether the next dollar belongs on Meta at all. Nobody measured on account performance can be expected to resolve those; they are not in scope. If the last three quarters produced a better ad account and a flat business, more execution capacity will not change the pattern.
With a fashion apparel and accessories brand, sales rose 249% and net profit 205% year over year with spend scaled 4x. New customer cost per acquisition and new-customer ROAS were tracked throughout, and reporting ran on blended sales and margin for the ownership team.
Run this checklist against whoever you are considering, including Plaid Testing. The free thirty minute audit exists for exactly that: tracking, account structure, three fixes, nothing to sign.
