What we believe about performance marketing.

The standard agency model is structurally rigged against the people paying for it. Retainers reward headcount, not outcomes. Statements of work get written to be defensible in a quarterly review, not to be honest about what works. The senior strategists who win the pitch disappear the week after the contract is signed, and the account is quietly handed to a team that is two years into the job.

This is not a moral failing of the people who work in agencies. Most of them are talented, careful, and trying to do the right thing. The problem is the incentive structure. When you bill by the hour, you are paid more for inefficiency. When you sell retainer, you are paid the same whether the work moved the number or not. When you scale by adding bodies, you are pulled toward the kinds of accounts that can absorb bodies, not the kinds of accounts that need someone senior thinking carefully.

Watch what happens when an agency hits a hard quarter. Headcount stays. The bar on new business drops. Account managers are told to expand existing engagements regardless of whether the existing engagements are working. The math of running a service business with high fixed costs forces every agency over a certain size to behave this way, no matter how well-intentioned the founders were when they started.

If your last agency disappointed you, the agency was probably not the problem. The model was.

Impressions, reach, engagement rate, share of voice, top-of-funnel lift. These are the metrics that get reported when there is no business outcome to report. They exist to fill a slide. They are not measuring what you care about. They are measuring what is easy to measure when the campaign did not work.

There are exactly three numbers that matter for most performance accounts: cost to acquire a customer, the value of that customer over time, and the rate at which you are willing to spend to acquire the next one. Everything else is a leading indicator that may or may not actually lead anywhere. If your reporting cadence does not start with those three numbers and end with what we are doing this week to move them, your reporting cadence is theater.

The reason agencies report on noise is that noise is reliably positive. Impressions almost always go up when you spend more. Engagement rate trends in your favor when you cherry-pick the audience. Brand lift studies are designed by the platforms that sell the inventory. None of this tells you whether the business made more money than it spent.

There is a second-order problem buried in this. When agencies report on noise, the operators they work for start to internalize noise as the goal. We have inherited accounts where the previous team had spent twelve months optimizing toward a metric that had no relationship to revenue. The campaigns hit the target. The business did not grow. Everyone got a quarterly bonus.

We report on three numbers. If we cannot move them, we tell you.

The largest single driver of paid media performance is creative. Not targeting. Not bidding. Not platform allocation. The hook, the script, the visual, the call to action. That is where the variance lives. A great creative against a mediocre audience will outperform a mediocre creative against a perfect audience nine times out of ten.

Despite this, most agencies treat creative as a brand exercise that happens upstream of the media work. The brand team approves the campaign in a deck. The performance team runs whatever the brand team approved. When the campaign underperforms, the performance team is held responsible for distribution, and the creative is held above scrutiny because the brand team already signed off on it.

We work the other way around. Creative is briefed against the metric we are trying to move. We test concepts in structured rounds and let the data tell us what is working. We do not get attached to ads we love. We get attached to ads that sell. If a beautiful piece of work has half the conversion rate of an ugly piece of work, the ugly piece of work runs.

There is a useful exercise we run early in every engagement. We ask the client to show us their five best-performing ads from the last twelve months and their five worst. Almost every time, the best performers look nothing like what the brand team would have approved in a vacuum. The hooks are blunter. The pacing is faster. The on-screen text is uglier. This is not a coincidence. The work that sells is rarely the work that wins design awards.

Beautiful ads that don’t sell are a waste of everyone’s time and money.

The first conversation with most prospective clients ends in the same place: they want to spend more money on ads. They have a number in mind, often a multiple of what they spend now, and they want a partner who can deploy that spend efficiently.

The honest answer, in the majority of those conversations, is that the account does not need more spend. It needs better tracking, a stricter conversion definition, a real attribution layer, and a creative testing process. Until those four things are in place, more spend will not produce a proportional lift in revenue. It will produce a proportional lift in waste.

We have run audits on accounts spending six figures a month and found that thirty percent of the spend was being credited to channels that did not actually drive the conversion. We have found tracking that double-counted purchases on iOS. We have found creative testing pipelines that ran one ad variant for six months because no one was responsible for replacing it. None of these problems get fixed by spending more.

The reason these problems persist is that fixing them is unglamorous. Server-side tagging audits do not produce a quarterly campaign launch. A stricter conversion definition usually makes the dashboard numbers look worse before they look better. A real attribution layer often reveals that one of the agency’s preferred channels is contributing less than the agency claimed. None of this is fun to present in a status meeting. All of it is necessary before the next dollar of spend can be deployed responsibly.

Spend is a multiplier on the infrastructure underneath it. If the infrastructure is broken, spend multiplies the brokenness.

We say no a lot. We say no to clients whose unit economics will not support performance marketing at the scale they need. We say no to projects we do not believe we can move. We say no to creative we know will not perform, even when the client loves it. We say no to channel mixes that look diversified on paper but dilute the spend below the threshold where any one channel can produce signal.

This is unusual. Most agencies are structurally incapable of saying no, because saying no costs them revenue. They are paid to nod, to accommodate, to find a way to make whatever the client asked for work. Their incentive is to keep the relationship going. Ours is to make the work compound, which sometimes means telling the client that what they asked for is wrong.

There is no version of high-performing work where the practitioner is not allowed to push back on the client. The client is not always right. The client is often optimistic, anchored to last quarter’s plan, or being pulled by a stakeholder who does not have the data we have. Our job is to bring the data and the experience and to be honest about what we see, even when it is uncomfortable.

Saying no in the first conversation is the cheapest version of saying no. Saying no in the third quarter, after the relationship has been built and the spend has been scaled and the team has been trained, is the most expensive. The agencies that cannot say no early end up saying nothing late, and the work atrophies under the weight of accumulated yeses.

If you want a yes-man, the rest of the industry will be delighted to take your money.

Performance marketing is not a series of campaigns. It is a compounding system. Every test that runs feeds the next test. Every dollar that goes through the system makes the next dollar smarter. Every creative round either disqualifies a hypothesis or strengthens the one that is winning. Done well, the account gets more efficient month over month, not because the platforms get easier, but because the operator running the account gets more specific about what works.

This is why we don’t take six-month engagements. The first ninety days are spent building the foundation. The next ninety are spent learning what the account actually responds to. By month seven, we are starting to see the compounding effect, the same spend producing meaningfully more revenue than it did at the start. Pulling the cord at that point is leaving most of the value on the table.

It is also why we will not work with operators who treat marketing as a tap to be turned on and off in response to short-term cash flow. If the spend gets paused every quarter, the system never compounds. The account permanently underperforms its potential, and no amount of senior attention will fix that.

Compounding requires three things from the operator: budget that does not whipsaw, patience that does not break under one bad month, and willingness to let the work change shape as the data comes in. None of these are technical. All of them are temperamental. The accounts we have grown the most are not the ones with the largest budgets. They are the ones run by operators who understood that the first six months of careful infrastructure work would be worth more than any short-term win we could have manufactured.

The work compounds when it is allowed to. Our job is to make sure it is.