Your Media Team Will Hit Every Target You Set. That's the Problem.

I inherited an account a while back where the media team had hit every number on their scorecard for six straight months.

Ad volume, on target. Hook rate, above benchmark. Average creative age, under the threshold. Nothing running below the ad level ROAS floor. Every square green.

Revenue had been flat for five of those six months and contribution margin was down.

Nobody was slacking. Nobody was lying. The team had done precisely what they were asked to do, and the account had gotten worse because of it.

The scorecard was not measuring the account. The scorecard was producing the account.

Why a Team Can Hit Every KPI and Still Lose Money

Charles Goodhart wrote a paper in 1975 on UK monetary policy that produced the cleanest mental model in operations. When a measure becomes a target, it stops being a good measure.

The standard illustration is the old parable of the nail factory. Told to produce more nails, it produces enormous quantities of tiny useless ones. Told instead to produce more weight of nails, it produces a handful of oversized useless ones. The measurement was never the problem. Making it the goal was.

Your ad account is full of nail factories.

The mechanism is not laziness or bad faith. It is that any number simple enough to put on a scorecard is also simple enough to satisfy without doing the underlying work, and a competent team under pressure will find that path whether or not anyone intends it.

This matters more in performance marketing than in most functions, because almost every metric in the platform is derivative. ROAS is a sensor for revenue, revenue is a sensor for whether you are creating something people want at a cost below what they will pay. By the time you are looking at hook rate, you are four abstractions from the thing that actually pays salaries.

The Difference Between a Sensor and a Target

Every metric in your account is one of two things.

A sensor tells you what is happening. You read it, you diagnose with it, you investigate when it moves. It informs decisions. It is not itself a decision.

A target is a number someone is accountable for hitting. It shapes behavior whether you want it to or not, because the moment a person's performance review touches a number, that number starts influencing every choice they make.

Most agencies and in house teams have no explicit line between the two. Metrics drift onto the scorecard because someone found them useful in a diagnosis once, and nobody asked whether making them accountable would produce the behavior they wanted.

Sensors inform. Targets steer. Swapping them is how good operators break good accounts.

The Five Targets That Quietly Break Ad Accounts

These are the ones I see most often, and each of them started life as a genuinely useful sensor.

Take the hook rate one, because it is the most seductive. Hook rate is a good sensor. If a creative is underperforming and the hook rate is weak, you have found your problem in ten seconds. That is real diagnostic value.

Now make it a target. Brief creators to hit 30% three second view rate. They will hit it, because hooking attention is an easy problem when you are allowed to hook the wrong attention. Your hook rate improves and your conversion rate falls, because the people you stopped were never buyers.

The metric did not lie. It answered the question you asked. You just asked the wrong one.

The last row deserves its own note. Holding a media buyer accountable for account activity produces a buyer who touches the account constantly, and constant intervention against normal variance is one of the most reliable ways to degrade performance. The job is making the right decision when one is needed, not making decisions as a default.

The Gaming Test

Before any metric goes on a scorecard, I run it through one question.

If someone improved this number by gaming it, would the headline metric actually improve?

Not "would they cheat." Assume good faith and assume normal human incentive response. If a competent person optimized aggressively and exclusively for this number, does the business get better or worse?

That single question sorts almost everything.

Notice the pattern in the safe column. Every metric that survives the test is one where the shortest path to improving the number happens to run directly through work you wanted done anyway.

That is the whole design principle. A good target is one where gaming it and doing the job correctly are the same activity.

Why This Hits Winning Accounts Hardest

Here is the part that is counterintuitive, and it is the reason this problem persists inside teams that are otherwise very good.

Plenty of sub-KPIs in an excellent account look worse than theoretical best practice. That is the nature of portfolio optimization. You cannot have every component be best in class and also have the whole system be best in class. Real portfolios have weak corners, and the weak corners are frequently load bearing.

Ad level last click CPA inside a consolidated campaign structure will look ugly for plenty of genuine winners, because the algorithm is optimizing against something broader than any single attribution model can see. Pause those ads to clean up ad level CPA and you are removing inputs the system was using. Headline performance flattens, and nothing on the scorecard explains why.

This is why the problem is worse in mature accounts than in broken ones. In a broken account, cleaning up the obvious losers helps. In an account that is working, the obvious losers are frequently part of how it works.

Which means the scorecard that helped you fix a bad account will actively damage a good one if you never revisit it. I have watched this specific transition happen more than once, and it is almost never diagnosed correctly, because everything on the report is green.

How I Actually Structure Targets

Three layers. Different numbers, different accountability, different review cadence.

Layer 1: One headline metric

One. Not three.

For most brands this is contribution margin in dollars, or new customer acquisition efficiency, or blended MER, depending on the stage and the constraint. A pre profitability brand with runway is optimizing for something different than an eight figure brand protecting margin.

The requirement is that it survives the gaming test and that everyone on the team can name it without checking. If your media buyer, your creative lead, and your analyst give three different answers to "what are we actually optimizing for," you do not have a headline metric. You have a dashboard.

Layer 2: Two or three guardrails

Guardrails are not goals. They are constraints on how you are permitted to pursue the headline.

Typically one on efficiency and one on cash. New customer CPA below a ceiling. Blended MER above a floor. Occasionally a third on mix, like new customer share of revenue, when the business is at risk of quietly becoming a retention business without noticing.

Guardrails bind, meaning whichever one is tightest governs the decision, but nobody gets rewarded for over performing on a guardrail. Beating your CPA ceiling by 40% while under spending is not a win. It is a missed opportunity dressed as discipline.

Layer 3: Input targets

The small number of leading metrics that passed the gaming test. Distinct angles tested per month. Cost per proven concept. New winner rate. These are the ones a creative and media team can actually control on a weekly basis, and they are the only place I put activity based accountability.

Everything else in the account is a sensor. Reviewed weekly, discussed in diagnosis, never assigned as a number anyone is judged against.

A Worked Example

The restructure I ran on the account I opened this piece with. Same team, same budget.

Before:

After:

Ad output dropped from 40 a month to 26. Angles tested went from 2 to 6. Average creative age went up, because profitable ads stopped getting paused for being old.

Contribution margin improved inside two months, and the improvement held. The team was doing less work and more thinking, which is usually what a scorecard problem looks like from the inside once you fix it.

The uncomfortable part of that meeting was telling a team that had been green for six months that half their targets were being removed. Expect that conversation. It goes better if you frame it as taking work away rather than as a correction, because it genuinely is.

The Questions I Ask Before Setting Any Target

1. Does it survive the gaming test? If gaming it does not improve the headline, it is a sensor. No exceptions, including for metrics you personally like.
2. Can the person being held to it actually control it? Holding a media buyer to blended MER when the site converts badly is not accountability. It is blame.
3. What is the shortest path to hitting this number? Write it down literally. If that path is not work you want done, you have your answer.
4. What behavior does this remove? Every target crowds something out. Usually it crowds out judgment, which is the thing you hired for.
5. Would a great operator's account fail this target? If yes, the target is wrong, not the operator.
6. How many targets is this person carrying now? Past about five, they stop functioning as priorities and start functioning as a checklist, which produces compliance rather than performance.
7. When does this expire? Targets that made sense during a fix rarely make sense once the fix worked. Review the scorecard quarterly with the same seriousness you review the account.

FAQ

What is the best KPI for a media buyer?

There is no single one, and the question itself is the trap. A media buyer should carry one headline metric tied to business outcome, two or three guardrails, and two or three input targets they can directly control. Anything beyond that becomes a checklist.

Should I set a hook rate target for my creative team?

No. Use hook rate as a diagnostic when a creative underperforms. As a target it produces openings optimized for attention rather than for the right attention, and conversion rate absorbs the cost.

Is ad volume a bad metric?

Ad volume is a useful sensor and a poor target. If you want to hold a team accountable for production, use distinct angles tested rather than assets shipped. Angles cannot be padded the way asset count can.

How many KPIs should a performance marketing team have?

In my experience, one headline, two or three guardrails, and no more than three input targets. Around five to seven accountable numbers total. Past that, people optimize whichever is easiest rather than whichever matters.

What is Goodhart's Law in marketing?

It is the principle that a measure ceases to be a good measure once it becomes a target. In marketing it shows up whenever a diagnostic metric gets promoted onto a scorecard and the team starts optimizing the metric rather than the outcome it was meant to indicate.

How do I know if a KPI has become a target problem?

Two signals. The metric improves while the headline stagnates, and the team can explain how they hit the number without referencing anything about the business. If the explanation is entirely mechanical, you have your answer.

Should ad level ROAS floors ever be used?

As a sensor, yes. As an automatic pause rule inside a consolidated campaign structure, rarely, because ads that look weak on a single attribution view are frequently contributing to how the system learns. Diagnose before you cut.

How often should I revisit the scorecard?

Quarterly at minimum. Targets are usually built to fix a specific problem, and they tend to outlive the problem by a long way. A scorecard that helped repair a struggling account will actively damage the same account once it is healthy.

Closing

The instinct when performance is soft is to add measurement. More granularity, more thresholds, more accountability. It feels like management.

What it usually produces is a team optimizing a scoreboard while the business underneath it drifts, and the reason it takes so long to catch is that the reporting looks better every month it gets worse.

The real job has not changed. Make something valuable, get it in front of people who want it, and do that for less than they will pay. Everything inside the ad platform is instrumentation in service of that. The dashboard is only worth having to the extent it makes you smarter about the actual work.

Sensors sense. You steer. The moment you hand the steering wheel to a sensor, you have stopped managing an account and started managing a number.

For the adjacent problem of evaluating a media buyer's actual judgment rather than their outputs, see how to audit a media buyer without micromanaging and the weekly dashboard we run across 300+ brand accounts.

Keep reading

Pieces I've written on related topics that pair well with this one:

• ROAS Is Not the Goal. It's a Sensor on the Dashboard., why the scoreboard metrics mislead the moment you optimize for them.
• The North Star Metric Trap: Why One Number Destroys Growth, the three layer metric architecture that works.
• How to Audit a Media Buyer Without Micromanaging, the four layer framework for evaluating judgment.
• The Weekly Dashboard We Run Across 300+ Brand Accounts, the nine metrics and their flag thresholds.
• 12 Metrics That Matter More Than ROAS for DTC Brands, the leading indicators worth watching.