Situation. An SMB-focused SaaS business, bootstrapped, growing fast, with early success on satellite radio that suggested broadcast radio could scale. Every marketing dollar had to work.

The number nobody had computed. How many impressions it takes to produce one website visitor. Roughly 2,000 to 4,000, as it turned out — a constant that decides in advance whether a media test can produce a readable result.

The test and how we made it readable. The first round of tests ignored this, and it cost $200,000 across several markets at $25K–$75K each. Working it through afterward: a $25,000 buy at local radio CPMs bought about two million impressions over four weeks. At the observed impression-to-visitor rate, that is roughly 1,000 visitors a month, or about 50 per weekday — against an existing baseline of 5,000 daily visitors. The tests were not inconclusive. They were undetectable, and they were undetectable before a dollar was spent.

The corrective was not a better read on the same data. It was a differently sized test: budget concentrated into fewer markets and a shorter window, scaled so that if the channel worked, the result would clear the noise floor.

Result. Broadcast radio became a primary acquisition channel. Annual customer acquisition volume rose fourfold, revenue grew 20–30% annually on a sustained basis, and the business was eventually acquired for over $170 million at 6.5 times revenue.

What transfers. A test that is too small does not produce a wrong answer — it produces no answer, and the channel gets written off on evidence that never existed. Required scale is calculable in advance from the client's own traffic and the channel's impression constant. If the budget will not support a readable test, the correct move is to defer it or concentrate it, not to run a diluted version and hope.

‍