One Question That Exposes Any Growth Agency: Who Owns the Forecast When It Misses?

It's the second Tuesday of the month. The agency's on the Zoom, screen shared, walking you through last month. Spend was up. ROAS held. There's a slide with a green arrow. Everyone nods. And not once does anyone open the number that actually pays your staff, because nobody in that room is on the hook for it.

I've sat in that meeting from both sides now, and that green-arrow slide is where most agency relationships quietly go wrong. Not with a blow-up. With a recap dressed up as a result.

So here's the one question I'd put to any growth agency before you sign, the one that sorts the operators from the order-takers:

When the forecast misses, who owns it?

Not who built the forecast. Anyone can build a forecast. Who's accountable when the month comes in under it, and what do they do at 9am the next day.

Most agencies have never been asked this. You can tell, because the good ones light up and the rest start talking about "factors outside our control." Below are the five questions I'd actually walk in with, and what a real answer sounds like for each.

1. Do you forecast at all, and how often do you revisit it?

The first filter is whether there's a number to miss in the first place.

Plenty of agencies don't forecast. They run ads, they report, they wait for you to be unhappy. There's no commitment on the table because committing to a number is scary, and reporting on the past is safe.

What I'd want to hear: a monthly forecast, revisited weekly. Not a static figure pulled in January and never touched. Markets move. Meta has an inefficient fortnight. Your supplier slips. A forecast that doesn't get re-cut every week isn't a forecast, it's a wish you wrote down once.

I believe the cadence matters more than the accuracy. An agency that re-forecasts every Monday is an agency that's watching. One that set a number in January and never mentions it again has stopped looking.

2. Can you show me contribution margin daily, or only ROAS monthly?

This is a big one. ROAS is the metric agencies love because it makes them look busy. Contribution margin is the metric that tells you whether you're actually building anything.

Here's the thing. You can hold a beautiful blended ROAS and still go backwards, because ROAS doesn't know your COGS, your shipping, your discounting, or the fact that half this month's profit came off returning customers who'd have bought anyway.

To put this into perspective: I've seen a homewares brand sitting at a 3.1x blended ROAS that everyone was happy with, while contribution margin after product and fulfilment was barely positive on new customers. The agency was reporting a number that felt great and meant almost nothing. The day we started tracking margin daily instead of ROAS monthly, the whole strategy changed.

What I'd want to hear: margin measured daily, not reconstructed in a spreadsheet six weeks later. If an agency can only tell you whether last month was good once last month is long gone, they can't steer. They can only autopsy.

3. When we fall behind, what's the actual playbook?

You will fall behind. Every brand does, in some month. The question isn't whether it happens, it's what the agency does in the 48 hours after.

This is the question that exposes whether there's a brain behind the account or just a hand on the budget button. Ask them to walk you through a real month where the forecast slipped. What did they pull? In what order? Why?

A good answer is specific and a little boring. Something like: "First we check whether it's a creative fatigue problem or a demand problem, because the fix is completely different. If new-customer cost is climbing but the active customer file is healthy, we lean into retention and pull back acquisition spend rather than torch margin chasing strangers. If it's creative, we've got the next three concepts already shot."

A bad answer is "we'd optimise the campaigns" or "we'd test new audiences." That's not a playbook. That's a vibe.

What I'm really listening for is whether they think in levers, and whether they know which lever moves which number. An agency that can't name its levers will just spend more when things get hard, which is exactly the wrong move and the most common one.

4. Do you think like our finance team or just our media buyer?

The best shift I've watched happen across this industry over the last few years is agencies moving from "how much can we spend" to "how much profit can we protect." The ones who made that jump treat marketing and finance as the same conversation. The ones who didn't are still optimising for a number on the ads dashboard that your accountant has never once looked at.

You want an agency that knows the difference between growth mode and monetisation mode and tells you straight which one you're in. That'll happily recommend you spend less this month if spending less makes you more. That talks about your blended numbers and your cash position, not just the slice of the world Meta can see.

What I'd want to hear: them asking about your margins, your AOV, your returning-customer rate, before they ever talk about creative or budget. If the first thing an agency wants is access to your ad account and not your P&L, you've learned something already.

5. What are you actually charging me to do that Meta no longer does?

This is the uncomfortable one, and the most important right now.

Meta automates the button-pushing now. Broad targeting, automatic placements, the algorithm finding your buyer. A lot of what agencies historically charged for, the manual campaign tinkering, the audience-stacking, the daily bid nudging, the machine does that, often better than a person fiddling at midnight.

So if an agency is selling you button-pushing, you're paying a premium for labour the platform gives away. That's a real question to put to them: what's left that's worth my money?

Here's my take on what's actually worth paying for in 2025, because it's the stuff a machine still can't do:

• Judgement. Knowing which lever to pull when the forecast slips, and having the nerve to recommend pulling spend back when everyone's instinct is to push harder.
• Creative strategy. The angle, the hook, the thing that makes a stranger stop. Meta will distribute your creative brilliantly. It won't tell you what to say.
• Accountability for the number. A throat to grab when the month misses, and a plan that doesn't start with "factors outside our control."
• The finance lens. Someone watching contribution margin every day so you find out you're bleeding on a Tuesday, not in next quarter's review.

That's the work now. Everything else is increasingly a commodity, and you shouldn't pay agency rates for a commodity.

What a good answer actually sounds like

If you boil all five down, you're really testing for one thing. Does this agency feel responsible for making your forecast come true, or just for explaining why it didn't?

That's the whole difference. Building a forecast is a spreadsheet exercise. Making it come true means watching margin daily, re-cutting the number weekly, knowing your levers cold, and owning the miss when it comes. One is a service. The other is a partner.

And the tell is almost always in how they talk about a bad month. The order-takers reach for the weather, the algorithm, the macro. The operators reach for the playbook.

Where to from here

You don't need to wait for a sales call to run this. Pull up your last three monthly reports right now and ask the simplest version of the question: when one of those months came in soft, did anyone own it, or did the recap just quietly move on?

If you can't find the moment where someone said "that's on us, here's the fix," that's worth sitting with. A quiet look over your account, your margins, and how your current setup actually responds when a month slips will usually tell you in an afternoon whether you've hired a partner or rented a pair of hands. If you'd like that look, that's exactly what a Signal/Noise Audit is for.