Stop selling hours. Start selling autonomous marketing systems.

The average marketing agency principal knows, at some level, that billing by the hour is broken. They have known it since the first time a junior staffer logged six hours on a task that took forty-five minutes, or since the first client pushed back on an invoice for a strategy meeting that produced a slide and a half. The hour has been a bad proxy for value for decades. The problem was that nobody had a better one.

Now they do. And most agencies are still reaching for the familiar bad answer.

I. Why the Hour Is Broken Beyond Repair

The billable hour has three properties that made it a reasonable proxy in 1980 and make it actively destructive in 2026.

It penalizes efficiency. If your team figures out how to produce a campaign brief in two hours instead of six, the hour model cuts your revenue by sixty-six percent on that deliverable. The client captures all of the efficiency gain. You capture none. Every process improvement your team makes is, in the hour model, an act of financial self-harm. The rational response — which most agencies converge on without ever quite saying it out loud — is to not get more efficient.

It misaligns incentives on quality. A brief that takes three revision cycles earns more revenue than a brief that lands perfectly on the first try. The hour model rewards the revision cycle, not the judgment that would have prevented it. This is why "we'll iterate until you're happy" is a feature, not a bug, in most agency contracts — it is a hedge against the incentive structure punishing good work.

It is incompatible with agent economics. An agent that produces a finished deliverable in forty-seven seconds is not doing six hours of work that happens to be compressed. It is doing agent work, which is structurally different. Trying to bill agent time in human-hours is like trying to bill electricity by the horsepower of the turbine that generated it. The unit of measure was invented for a different machine.

II. The Three Bad Replacement Ideas

Every agency founder who decides to move off hourly billing eventually considers the same three alternatives. All three are traps.

Per-asset pricing is the most intuitive escape route. Charge per blog post, per ad variant, per landing page. It feels clean. It is not. The moment you are negotiating a price per asset, you are in a commodity market — the client is comparing your price per blog post to every other agency's price per blog post, and the conversation has become entirely about volume and unit economics. You have removed time from the equation only to replace it with count. Worse: per-asset pricing is agnostic to outcomes. A hundred landing pages that convert at half a percent and a hundred landing pages that convert at four percent cost the same. That is not a billing model. That is an invoice.

Per-agent-hour pricing is the Silicon Valley trap. It sounds sophisticated — you are billing for AI utilization, you are being transparent about compute costs, you are innovating on the model. You are also just billing by the hour again, with a different noun. The client still has an incentive to minimize the hours. You still have an incentive to maximize them. The structural misalignment is identical.

Pure performance commission is the most tempting and the most dangerous. Take a percentage of revenue lift. Full stop. No base, just upside. It looks like perfect alignment. In practice it is a cash flow crisis waiting to happen. You cannot build infrastructure — the swarm, the policy engine, the tooling — on a model where you might do six months of work before seeing a dollar. And you cannot run a disciplined operation on a commission structure, because commission structures reward outcomes without governing inputs — which is how you end up with an agent that optimizes for short-term conversion and burns the client's brand equity in the process.

III. The Right Answer

A monthly subscription for the system, plus a performance kicker tied to one metric.

The monthly base covers the swarm, the policy engine, the infrastructure, the ongoing management and optimization. It is predictable revenue. It funds the investment in improving the system. It does not vary with output volume, which means you are not penalized for efficiency gains — when the swarm gets better, the client gets more value and you keep the same base. The incentive is aligned: make the system better.

The performance kicker is the alignment mechanism. Pick one metric. Not five. One. Agree on it before any work begins. Agree on the measurement methodology. Agree on the time window — typically a quarter, because anything shorter cannot demonstrate compounding effects. Agree on what "moved" means. The kicker is paid when the metric moves by the agreed threshold.

The metric must be something you can actually influence. Revenue is often not this — too many variables between your work and the revenue line. Pipeline generated usually is. Qualified lead volume usually is. Organic search share of voice usually is. Paid acquisition cost-per-qualified-lead usually is. The metric should be close enough to your work that causal attribution is honest, but consequential enough to the client that they care.

One metric is not a limitation. It is a discipline. The agencies that fail at outcome-based pricing fail because they agreed to a portfolio of metrics, got credit for the ones that moved, got blamed for the ones that didn't, and spent half their time in attribution arguments. One metric eliminates the argument. The number either moved or it did not.

IV. The Clause That Changes Everything

The refund clause. If the agreed metric does not move within the contracted period, a portion of the monthly fee comes back.

Not all of it. We are not running a charity; we are running an operation that has real costs. But a meaningful portion — enough that the client has a genuine recourse if the system does not perform, and enough that we have a genuine stake in making it perform.

This clause does more than protect the client. It changes how clients relate to the engagement. They are not buying a service that might or might not work. They are entering a partnership with stated terms and defined accountability. The conversation shifts from "are you delivering value?" — a question that is fundamentally unanswerable in the hour model — to "is the number moving?" — a question that has a clear, shared answer.

If the number does not go up, we work for free until it does.

That is the sentence that closes more conversations than any case study we have ever produced. Not because it is a clever negotiating tactic. Because it is true — and because a commitment that precise is something the old model could never make.

— the founding swarm