I Fired My Best-Performing Marketing Channel. Here's Why.

Last spring I killed the marketing channel that was bringing in 34% of new Happierleads signups. On paper it looked insane. The dashboard showed it was our best-performing acquisition source by volume. The CAC was reasonable. The signup graph was going up and to the right.

I killed it anyway. Sixty days later the business was measurably healthier. Revenue was up, churn was down, and I had a chunk of budget and attention back.

Here's the part that took me a year to figure out: the dashboard was lying. Or more accurately, the dashboard was answering the wrong question.

TLDR

• A channel doing 34% of signups was doing 6% of paid conversions. The volume was cosmetic.
• Signup-weighted attribution flatters bad channels and buries good ones.
• The only attribution metric I trust now is revenue at day 90, cohorted by first-touch source.
• Killing the channel freed budget for a smaller source that was quietly doing the real work.
• Most founders are optimizing the wrong number and don't know it.

The channel that looked great

It was a directory-style listing site. Paid placement, decent traffic, a steady drip of signups every week. The CAC per signup came in around $12, which felt like a bargain against our other channels sitting closer to $40.

For about eight months I patted myself on the back every time I looked at that row of the spreadsheet. Cheap signups. Growing volume. What's not to love.

Then I ran a cohort analysis I'd been putting off. Grouped every signup from the previous year by first-touch source, and instead of asking "did they sign up," I asked "were they still paying at day 90."

The number that changed my mind

Signups from that channel converted to paid at about a fifth of the rate of signups from organic search. And the ones who did convert churned within two months at nearly triple the rate of every other source.

CAC per signup was $12. CAC per retained customer was north of $600. Against a channel where I could get a retained customer for about $180 through content.

The buyers coming from that directory were tire-kickers. People shopping software the way you shop socks on Amazon. They signed up, poked at the product for a day, never installed the tracking, and disappeared. Great for the top of the funnel graph. Terrible for the business.

Why the dashboard fooled me for eight months

Every analytics tool defaults to signup-weighted attribution because signups are the easiest event to count. First-touch, last-touch, multi-touch — they all lean on the same top-of-funnel event as the currency of "performance."

But a signup isn't a customer. A signup is a stated intent, and stated intent is one of the cheapest things in B2B. What matters is whether that person shows up in your dashboard six weeks later still paying. That's the only marketing metric I care about anymore.

If your attribution stops at signup, you're measuring how good you are at attracting curiosity. That's a different job than growing a business.

What replaced it

I took the budget and put half of it into deeper content on the pages that were already converting. The other half went into something almost nobody thinks of as a channel — reading Happierleads' own visitor data and reaching out to companies who were repeat-visiting our pricing and comparison pages. Small volume. Absurd conversion. The retained-CAC on that motion came in under $90.

This isn't a plug. It's the honest answer to what worked. The people already researching you convert at rates that make paid directories look like a joke, and most companies leave that traffic anonymous because they don't bother identifying it.

The exercise worth an afternoon

Pull every signup from the last 12 months. Tag each one by first-touch source. Then filter to only the ones still paying at day 90. Recalculate CAC on that filtered set.

The channel rankings will scramble. I promise. One of your "top" channels will collapse. Something you barely think about will move up. And you'll spend the next quarter reallocating budget based on what's actually building the business instead of what's decorating the dashboard.

I sat on that directory bill for eight months because the top-line number was flattering. That's the trap. Vanity attribution is comfortable. The cohort math is uncomfortable and correct.

Talk next week,
— George