A South African insurance brand came to us with a warranty campaign that had been underperforming for months.
The budget was right. The creative was polished. The landing page converted. The targeting looked like every other insurance campaign running on Meta. And yet — leads were expensive, quality was low, and the campaign had been limping along at mediocre cost-per-lead numbers since launch.
The obvious answer was to change the ad.
We told them not to.
The Default Reflex Is Wrong
When a paid media campaign underperforms, the agency instinct kicks in almost immediately: refresh the creative, test a new hook, try a different CTA. More creative, more often. A/B test your way out of a bad result.
Sometimes that’s the right call. But it assumes the ad itself is the problem. And in this case, we didn’t think it was.
The creative was talking to the right pain. The offer was competitive. When we ran the landing page through a proper conversion audit, it held up. So if everything in the funnel was working, why were leads costing what they were costing?
We went back to the targeting.
What the Audience Actually Looked Like
The campaign was targeting vehicle owners in-market for “car insurance.” That’s a reasonable starting point — it’s the standard audience for this category on Meta. But warranty cover and standard car insurance attract very different buyers, and bundling them into the same targeting pool creates a mismatch.
Car insurance is a near-universal need. Everyone with a vehicle has to have it. The audience is broad, the competition is dense, and the CPM reflects that. You’re fighting every other insurer for the same pool of in-market consumers.
Warranty cover is different. It’s not mandatory. It’s an active choice made by someone who’s worried about repair costs on a vehicle that may no longer be under manufacturer warranty. That’s a more specific state of mind — and it tends to cluster in a more specific demographic.
Our hypothesis: the people who have the most urgent need for warranty cover are those driving older vehicles. Six-plus years old. Probably out of factory warranty. Probably staring down a service bill they didn’t plan for. Lower premium sensitivity, higher intent.
That person is in the audience the campaign was running to — but they’re buried in it. And the creative wasn’t speaking to them specifically.
The Hypothesis We Proposed
We proposed splitting the campaign into a new test audience: vehicle owners aged 35–55, targeting signals consistent with longer-term vehicle ownership, and creative that named the specific moment of pain.
Not “protect your vehicle” — generic, competes with everyone.
Instead: “Your car’s warranty expired. Here’s what happens next.”
That framing does something precise. It speaks directly to the person who knows their warranty has lapsed and has been quietly aware of the exposure. It doesn’t need to explain the problem. It just names it. Anyone not in that situation scrolls past. Anyone who is — stops.
We also hypothesized that this audience would have lower premium shock. Someone buying comprehensive car insurance in a competitive market has been conditioned to compare premiums aggressively. Someone who needs warranty cover on a high-mileage vehicle isn’t in the same mindset. The conversation is different. The creative needs to match it.
Why We Didn’t Just Change the Ad
I want to be clear about what we didn’t do here. We didn’t walk into the debrief and say “the creative isn’t working, let’s test new hooks.” We didn’t recommend doubling the budget to brute-force better results. We didn’t offer five creative variations and suggest running them all simultaneously.
We built a hypothesis. One specific, testable idea about who the right audience was and why the current audience wasn’t converting at the expected rate. Then we set up a test designed to prove or disprove that idea.
That’s a different process than “keep testing until something works.” It’s slower to brief, harder to explain to a client, and produces less visible activity in the short term. But it’s how you actually learn something from a media test — and how you avoid spending the next quarter cycling through creative variations that all fail for the same underlying reason.
If we’re wrong about the audience hypothesis, we’ll know in four weeks and we’ll have learned something specific. If we’re right, we have a scalable, repeatable audience strategy for this product category.
What a Hypothesis-Driven Test Actually Looks Like
For this campaign, the test is structured around a single variable: audience definition. We’re not changing the creative significantly. We’re not changing the landing page. We’re not changing the budget allocation by a meaningful amount.
We’re running the same core message — tweaked for specificity — at a new audience segment, and comparing cost-per-lead and lead quality against the existing campaign over a four-week window.
Everything else stays constant. That’s the point. If you change the creative and the audience and the budget at the same time, you don’t know what moved the needle. You got a result, but you learned nothing you can apply systematically.
A hypothesis test has a clear structure: this is what we believe, this is why we believe it, this is how we’ll know if we’re right, and this is what we’ll do with the result either way.
The Honest Part
We don’t have the results yet.
The campaign went live in the last two weeks of April 2026. We’ll have meaningful data toward the end of May. We’re not publishing results we don’t have — and we’re not going to dress up the current CPL numbers as validation of the hypothesis before the test has run.
What we can say is this: the logic is sound. The audience distinction is real. And the framing we’ve taken to the creative — naming the specific moment of pain rather than leading with generic protection messaging — is backed by what we know about how people make decisions around financial products they weren’t planning to buy.
But hypothesis-driven work means being honest about uncertainty. We think we’re right. We’ll find out in four weeks.
When the results are in, we’ll publish the follow-up.
The Question You Should Ask About Any Underperforming Campaign
If you have a campaign that’s been running for more than six weeks with consistently disappointing results, the reflex question is: what do we change? That’s the wrong starting point.
The right question is: what do we believe is causing this, and how would we test that belief?
It’s a harder question to answer. It requires you to actually form a view — to commit to a diagnosis instead of just generating activity. But it’s the only question that produces learning instead of just output.
We kept running the existing campaign while we built the hypothesis. We didn’t blow up what was there. We just stopped pretending that cycling through creative was going to solve an audience problem.
If your current campaign is underperforming, don’t ask your agency to refresh the creative. Ask them what they think is wrong — and whether they can explain, specifically, why they think that.
We’ll update this with results in May.
