One image was costing us £348,000 per year, but we didn’t even know it.
We were running debt consolidation ads, spending around £100,000 a month for a client. The campaign was working. Leads were coming in. Sales were happening. By most measures, it was a success.
But I had a hunch we could do better.
The landing page featured a stock photo of a smiling family. It was fine. Professional. Exactly what you'd expect. And it had been running unchanged for months because "if it ain't broke, don't fix it."
I wanted to test something different. Our research told us these customers weren't really buying debt consolidation. They were buying peace of mind. Relief from the 3am ceiling-staring, wondering how they'd make it through the month. The smiling family felt aspirational, but it didn't speak to where they actually were.
So we tested a new image. Something calmer. Something that felt like the relief they were seeking, not the destination they couldn't quite believe in yet.
For two weeks, we split the traffic. Half saw the original. Half saw the new version. The test cost about £3,000 in ad spend to get statistically significant data.
The new image beat the original by 29% in terms of click to lead.
At their volume, that single change saved £29,000 per month. £348,000 per year. From one image. One test. One willingness to spend money finding out what worked better.
That £3,000 "cost" delivered a 116x return in the first year alone.
This is what paying to learn actually looks like.
Here's the uncomfortable truth about paid advertising:
You cannot think your way to what works. You have to buy your way there.
You can research. You can strategise. You can study competitors and build personas and craft what you're certain is the perfect campaign. But until you've spent money putting it in front of real people, you don't know anything.
You're guessing. Educated guessing, maybe. But guessing nonetheless.
The campaigns that actually work - the ones that generate leads consistently, profitably, at scale - are never the first version. They're version three, version five, version ten. They're the result of testing headlines that bombed, images that flopped, audiences that didn't convert, and offers that fell flat.
Every winning campaign is built on the corpses of losing tests.