Every October the same conversation happens in a few thousand founder inboxes at once. The 2027 budget is due, the agency line is the second-biggest number on the marketing page, and somebody asks the obvious question: do we keep paying this?
The question is right. The way most people answer it is not. There are three doors, and the one everyone is tempted by this year is usually the most expensive.
Door one: keep it
Renew the retainer, negotiate ten percent off, hope next year's pipeline looks different from this year's. Most companies take this door by default, because the switching cost feels bigger than the fee.
It's the right door in exactly one case: the agency is producing, you can trace the deals it sourced in your CRM, and the cost per closed deal is one you'd happily pay again. If you can say all three sentences out loud without hedging, stop reading and renew. If you can't, the fee isn't the problem. The structure is. You're paying for effort, and effort is what you've been getting.
Door two: cut the agency, keep the ads
This is the one we're hearing about most. Drop the agency, keep the Google and Meta spend, run it in-house with AI tools. It sounds like the modern answer, and sometimes it is.
But notice what it actually does to the risk. You've removed the one party whose job was to turn media into meetings, and you're now funding the media yourself, with nobody on the hook if it doesn't produce. You cut the fee and kept every dollar of the risk. Then you added some, because the person now running paid in-house is usually doing it as a fourth job.
Door two works when two things are true: someone on your team already knows how to run paid well, not just how to log in, and your spend is small enough that a bad quarter is a lesson rather than a crisis. For a company spending a few thousand a month, that's a reasonable bet. For a company spending $15,000 a month with a sales team waiting on the output, it's the same risk you had before, now with less expertise behind it.
Door three: stop paying for effort
This door didn't exist two years ago. Find a partner who builds the engine, funds the ad spend, and gets paid only from the new revenue that shows up. The fee disappears and the risk moves with it, off your P&L and onto theirs.
It exists now because AI collapsed what it costs to build and run a growth engine. When delivery is cheap, a firm can afford to bet on the outcome instead of billing for the hours. We wrote up the math behind that if you want it. The short version is that the old agency model couldn't take this bet without risking a payroll, and the new one can.
Door three is not for everyone, and anyone who tells you otherwise is selling door one in a costume. It needs a product that already sells, a team that closes the meetings it's handed, margins with room to share, and a market big enough that more pipeline moves the number. If you're pre-revenue or still finding product-market fit, a performance partner is a bad idea for both of you.
The honest version of each door
Keep it, if the deals are traceable and the cost per deal is one you'd pay again. Take it in-house, if you have a real operator and the spend is small. Move the risk, if you have a product that sells and a team that closes, and your bottleneck is at-bats.
Why the decision is due in October, not January
The budget is due in Q4, but that's not the real deadline. The real deadline is lead time.
A growth engine takes weeks to prove and a quarter to compound. Meetings booked in November close in January and February. Meetings booked in February close in April. If you make the decision when the budget resets in January, you've lost a quarter of pipeline waiting for a number on a spreadsheet to roll over. We wrote about that timing separately in January pipeline is built in November.
So pick the door now. If it's door three, we're taking a small number of companies into proof sprints before the holidays, and you can run your own numbers in thirty seconds before you talk to anyone.