Guide / Commercial decisions
Most brands don't have a bad agency problem. They have a bad selection process. You judged the last one on the pitch deck. If you judge the next one the same way, you'll be here again in eighteen months.
Reading time
11 minutes
Written for
Founders and marketing leads shortlisting Google Ads agencies, or about to. Not for agencies.
Companion guides
The premise
The pitch deck is a sales asset. It is not evidence.
Every agency in your shortlist has a deck with a 5x ROAS graph, a founder testimonial, and a case study that fits their argument. That is table stakes. It tells you nothing.
What you are trying to distinguish is not agency A from agency B. You are trying to distinguish a media buyer from a commercial partner.
Media buyers optimise the account you show them. Commercial partners ask why the account exists in that shape at all.
The six questions below force that distinction. If they can't hold up under them, no case study will save the engagement.
Read this twice
You are not hiring an agency. You are hiring an opinion about your P&L. If they don't have one before they've read it, they don't have one at all.
The scorecard
Ask every shortlisted agency the same six questions in the same order. Score each answer against the good/bad columns below. Anyone scoring below four is disqualified — no matter how good the deck was.
Question / 01
A good answer sounds like
They open with contribution margin, break-even ROAS by product group, and customer payback. They talk about the P&L before the platform.
Disqualify if you hear
They open with account structure, PMax asset groups, and 'quick wins'. The word 'margin' doesn't appear in the first hour.
Question / 02
A good answer sounds like
The first request is a SKU-level cost file — COGS, shipping, returns, margin. They won't quote a target ROAS without one.
Disqualify if you hear
They quote a target ROAS in the pitch deck. It's the same number they quoted the last five prospects.
Question / 03
A good answer sounds like
They ask what you're overstocked on, what's tying up cash, and what you cannot afford to sell out of. They treat ads as an inventory lever.
Disqualify if you hear
They treat every SKU as interchangeable. Inventory doesn't come up.
Question / 04
A good answer sounds like
Within two weeks of access they can point to structurally unprofitable SKUs and tell you what happens to the P&L if you exclude them.
Disqualify if you hear
Every SKU is a growth opportunity. Nothing is a loss leader. The word 'exclude' is not in the vocabulary.
Question / 05
A good answer sounds like
They can point to a scenario where the right answer is to reduce spend. They've done it for other clients and they can name the account.
Disqualify if you hear
Every recommendation is a growth recommendation. Scaling is the only outcome on the table.
Question / 06
A good answer sounds like
You keep the MCC. You keep the Merchant Center, GA4, Tag Manager, CSS. They are granted access — never the reverse.
Disqualify if you hear
The account sits in their MCC 'for efficiency'. The CSS is theirs. Leaving means rebuilding.
The disqualifiers
Not warning signs. Not concerns. Disqualifiers. If you see any of the six in the pitch or the contract, walk. No amount of case studies or testimonials makes up for a structural incentive problem.
Percent-of-spend pricing
Their fee grows when your spend grows. Which means their commercial incentive is not to recommend you spend less — even when spending less is the right answer. The conflict is structural and it never goes away.
The universal case study
The same three case studies presented to every prospect regardless of sector, size, or business model. If the pitch would work for a jewellery brand and a pet food brand and a SaaS company, it wasn't written for you.
A blended ROAS target with no cost input
They quote a target ROAS number before they've seen a cost file. That number is either a bluff or a lift from the pitch deck of the last brand they pitched. Both are disqualifying.
'We'll rebuild it in week one'
A confident restructure plan before they've opened your account. You're not hiring a diagnostician, you're hiring someone with a template. The template is the same for everyone.
The 12-month lock-in
A long tie-in with a heavy exit fee is a confidence signal — theirs, not yours. Confident agencies win the twelfth month by earning the eleventh. They don't need a contract to hold you.
Ownership sits with them
The MCC, the CSS, the tag container, the feed tool — all sit inside their infrastructure. The next switch will not be your decision to make cleanly. Fix this before you sign.
The eight questions
The questions matter less than the response time. If any of these produce a pause, a glance sideways, or a rehearsed answer, that's the signal.
01
Who owns the MCC, Merchant Center, GA4, GTM and CSS at the end of the engagement?
Answer must be: you. Anything else is a structural problem.
02
Show me an account where you recommended spending less. What happened?
If they can't name one, growth is the only lever they know how to pull.
03
How do you set a ROAS target for a new client?
The right answer starts with margin, returns and payback — not a benchmark.
04
What would you refuse to do for us?
An agency without a 'we won't do that' list is an agency that will do anything for the retainer.
05
How is your fee structured, and what changes it?
You want a fixed fee that doesn't move when spend moves. Otherwise the incentive drifts.
06
Can I speak to the client you fired?
Every good agency has one. If they've never ended a relationship, they've never held a standard.
07
Who will actually work on the account — and can I meet them today?
The pitch team is rarely the delivery team. Meet the delivery team before you sign.
08
What does your 90-day review look like?
If it opens with ROAS and revenue, you're back where you started. It should open with margin and payback.
The trial structure
You can eliminate 80% of the switching risk by structuring the first ninety days as a test, not a marriage. Any agency worth hiring will accept this structure — they'll expect it.
01
The paid audit
A short, paid engagement — not a free pitch. Fixed scope, fixed fee, fixed deliverable. You keep the output whether or not you sign the retainer. If they won't sell you a paid audit, they don't have a defensible product.
02
The 30-day pilot
Read-only access for two weeks, then a limited scope of live changes for two more. You judge them on what they saw that the last agency didn't — before they touch the structure.
03
The 30-day exit clause
The retainer has a 30-day rolling notice from day one. Not 90. Not 6 months. If they're good, they'll earn month two. If they're not, you're not trapped.
The contract
Read the contract before you read the pitch deck a second time. It tells you more about how the agency operates than the deck ever will.
Fixed fee, not percent-of-spend. Rolling notice, not annual tie-in. Data ownership stays with you. No IP clauses over campaign structure or ad copy. No auto-renewal buried in section nine.
If the contract is longer than a page, ask why. The answer is usually 'to protect us'. That is the wrong direction of protection.
A confident agency has a short contract, because they intend to earn month two.
If we're on your shortlist
Send us your cost file and read-only account access. We'll come back with a commercial read on what's inside the account and what we'd stop, start, and leave alone in the first ninety days. It's the same output every other shortlisted agency should be offering. Capped at four audits a month.
Request the auditWhat you get
A written commercial read on the account: the SKUs that lose money, the tracking that's misfiring, and where the P&L is leaking through the media plan.
What we ask for
A cost file. Read-only access to the ad account. Thirty minutes of your time.
What we don't do
Sales calls disguised as audits. Percent-of-spend fees. Long tie-ins. If we can't help, we'll tell you and point you somewhere that can.