The performance trap: When metrics start shaping mediocre ideas

Publishing runs on more inputs than most people give it credit for. The idea. The argument. The timing. The voice behind it. Whether anyone needed to read the thing in the first place.

Click-through rate is one input among many and a minor one at that, but somehow it’s the one running the whole show.

Sit in on most content meetings and CTR is the only number anyone defends with any real conviction.

Nobody’s asking “is this true” or “does this sound like us.”

They’re asking how many people opened the door. Nobody’s checking what happened once they walked through it.

That’s the problem this piece is about.

Metrics were meant to be one voice at the table. Somewhere along the way they took the chair, the microphone and the minutes, and everyone else let it happen because arguing with a number feels like arguing with the weather.

The fix isn’t throwing the number out. CTR is genuinely useful for what it measures.

The fix is putting it back in its lane. One input, doing one job, not running the department.

What metrics actually measure

CTR measures whether a headline and a thumbnail convinced someone to click. That’s it.

It tells you nothing about whether they read the piece, whether it changed how they thought about something, whether they forwarded it to a colleague with a note saying “read this.”

It tells you about the door, not the room.

The same applies to most content metrics:

• Time on page is shaped by font size and mobile formatting as much as quality.
• Bounce rate is ambiguous in ways that most dashboards don’t acknowledge.
• Social shares correlate with emotional valence and topic breadth, not depth or originality.

Every metric is a proxy for something and proxies have a way of replacing the thing they’re supposed to represent.

Goodhart’s Law is usually framed as an economics problem, but it describes content work with uncomfortable precision.

When a measure becomes a target, it ceases to be a good measure. You write for the metric, the metric improves, the work gets worse and nobody in the weekly reporting meeting has a way to address that.

The client problem

I worked with a fintech client recently who was, not unreasonably, obsessed with CTR.

Every content decision ran through it. Headlines were tested, adjusted, tested again.

Topics were chosen based on historical click data. The briefs arrived pre-filtered through what had performed before.

The content was fine. It was genuinely fine. It was well-structured, readable and consistently got a decent click rate.

It was also, over about six months, becoming indistinguishable from everything else in the category.

Not because anyone was doing bad work, but because the optimisation process had trimmed anything that made the brand sound like a specific organisation with a particular point of view.

This is the difficulty.

From the client’s perspective, the numbers looked good. They had no obvious instrument for measuring “our content used to feel like us and now it doesn’t.”

That’s not a metric. That’s an editorial judgment and editorial judgment is exactly the thing that feels unaccountable to people who’ve been told that accountability means data.

Creatives and strategists working in this environment are often trying to argue for things that don’t show up in a dashboard.

Reach in the sense of provoking a real conversation. Brand distinctiveness. The piece that doesn’t perform in the first week but builds slowly because it says something true in a way that nobody else in the category is saying.

These are real outcomes. They’re just invisible to the tools most clients are using.

How the voice goes

Voice in writing is partly about what you choose to say and partly about what you allow yourself not to say.

It lives in the aside, the small observation, the moment of honesty that isn’t strictly necessary but earns the reader’s trust.

Most of that stuff doesn’t click well. It’s not designed to.

It’s designed to make someone feel like there’s a person behind the work rather than a content production process.

CTR optimisation is fairly hostile to all of that.

You can feel when you create or edit content.

• The intros get bolder but blunter.
• The conclusions start driving harder toward an action.
• The middle sections, where the interesting thinking usually happens, get trimmed because they’re “not adding to the journey.”

Each of these edits is individually defensible. Together they hollow the piece out.

I’ve read enough high-CTR content to notice that a lot of it sounds the same regardless of brand, industry, or subject matter.

The cadence is identical. The paragraph breaks happen at the same intervals. The rhetorical moves repeat.

It’s the written equivalent of a building designed entirely around footfall data, technically functional, oddly alienating.

Solutions that actually work

You can’t entirely solve this from inside a metric-driven brief. But you can change the terms.

The first move is expanding the measurement frame.

CTR is a useful signal for distribution and headline quality, fine, but it shouldn’t be the only signal in use.

Qualitative feedback, reply rates to email content, whether sales teams are actually using the content in conversations, whether the piece got picked up or cited elsewhere.

These are all evidence of performance that most reporting structures ignore because they’re harder to pull into a spreadsheet.

Naming them explicitly, building them into how success gets defined, changes what gets made.

The second move is separating content types by purpose.

Not everything needs to click well. Some content is designed to build authority slowly, to say something the industry isn’t saying yet, to attract a particular kind of reader rather than the maximum number of readers.

Treating those pieces by CTR metrics is like judging a reference book by its cover art. It applies the wrong measure to the wrong object.

The third move, and this is the one that requires the most trust, is keeping some space in the content strategy that isn’t pre-justified by data.

A piece written from genuine curiosity or genuine disagreement, without a historical click benchmark to chase, occasionally produces something that the algorithm didn’t predict and the audience genuinely values.

You can’t schedule for it. You can create conditions where it’s allowed to happen.

The underlying problem

The reason this is hard isn’t that clients are wrong to care about metrics. They’re not.

It’s that metrics got there first, before the field had properly worked out what it was trying to measure and now they’re embedded in how content gets commissioned, reviewed and valued.

What good content actually does, shifts thinking, builds trust, gives a brand a recognisable sensibility, makes someone feel like they’ve encountered a real point of view rather than processed content.

These outcomes are real but they’re slow and they’re hard to attribute. Metrics are fast and they’re easy to attribute.

That asymmetry shapes everything.

The work is convincing people that the slow outcomes are worth measuring and that the instruments for measuring them are worth building.

It’s an easier sell once you can show someone the alternative, which is a content archive that clicks well and says nothing.

If that sounds familiar and you’d like to think through how to build a content approach that performs without losing its point of view, get in touch.