On 12 January I cancelled the only marketing channel we had. The invoice was $14,200 a month, the return was 0.8x, and I had spent nine months telling myself the graph would bend.

What the paid funnel actually returned Twelve months of spend produced 3,411 activated accounts and $186,000 of first-year revenue. The channel was not just unprofitable, it was replacing the work that would have made us profitable: the referral loop, the documentation, the two partnerships we kept postponing because "the ads are running."

I have the export. It is not a story about intuition.

The rebuilding, in weeks Weeks 1 to 6 were pure subtraction — I closed the campaigns, kept the analytics, and wrote down the four channels I believed could work with a team of three. Weeks 7 to 19 were bad. Revenue fell 31% and I ran the company on a cash buffer I had stopped topping up.

Week 20 was the first organic month above the paid baseline. Not by much: $8,400 versus $7,900.

What actually compounded Three things, in order of size: a public changelog that customers started linking to, two integration partners whose support docs sent us traffic with no contract, and a referral credit system that cost 11% of the revenue it produced.

None of those existed in month one. All of them required someone to do unglamorous work with no dashboard to watch.

The number I would put on the record Eleven months of zero ad spend produced $1.42 of revenue for every $1.00 the paid funnel had produced, at a third of the working hours. If I had tested that at month three instead of month nine, the company would be two years further along.

The spreadsheet is the receipt. The lesson is that the expensive channel was never the funnel — it was the excuse not to build the compounding one.