← marketing-mindset · free tool, no signup, nothing stored, runs in your browser
Which of your channels can actually fund a verdict?
A channel is not a strategy. It is a pipe with a volume, a price per contact, and a conversion rate you suspect is wrong. What makes it a channel you can act on is whether it can deliver enough contacts to one honest two-arm test before your patience runs out. That is arithmetic, not opinion: the contacts a strict test needs per arm (the floor), divided by the contacts one arm actually sees in a month (half of the channel's monthly volume), gives the months until a verdict.
The floor is the same number for every channel in the table, and that is the point. What you test is a rate, and a rate does not care where the contact came from. At a 3% base rate and a +20% relative lift, a strict two-proportion comparison at 80% power and alpha 0.05 needs 13,914 contacts per arm — 27,828 contacts in both arms. A channel that sends 9,000 contacts a month puts 4,500 in each arm and reaches that floor in four months. A channel that sends 1,200 a month puts 600 in each arm and needs twenty-four. Same floor, same test, different calendar.
Cost is what turns that calendar into a decision. One full two-arm test costs the floor per arm, times the cost per contact, times two. Between two channels that both finish inside your horizon, the cheaper test is not the better channel — it is the first test. Run it, read the verdict, then spend the same money on the next one with a real answer behind you instead of a hypothesis.
And if no channel finishes inside your horizon, the honest output is not a smaller test. Below the floor you are buying a sample, not a decision: the two arms will differ by noise and you will explain the noise. A channel that cannot finish one honest test inside your horizon is a subscription, not a test — it produces activity every month and a verdict never. The three ways out are named at the bottom of the result: accept a bigger lift, widen the horizon, or buy volume.
1. Your channels
One per line or comma separated, in the form name:contacts_per_month:cost_per_contact
— for example email:9000:0.02,linkedin:1200:0.35,seo:30000:0.005. Cost is whatever a contact
costs you in your own currency (a platform fee, a list price, a click price); it is printed, never converted.
2. What counts as worth detecting
| Setting | Value |
|---|---|
| Base rate — the conversion you are trying to move (0.03 = 3%) | |
| Relative lift worth detecting (0.2 = +20%) | |
| Months you are willing to wait for a verdict | |
| Power (0.8 or 0.9) | |
| Alpha (0.05, 0.01 or 0.1) |
3. Result
4. Plain text version
How to read the answer
The months column is not a prediction about how fast the channel converts — it is how long the channel takes to produce the contacts a verdict needs. One arm sees half the monthly volume because a strict test splits 50/50. Three things the table refuses to hide:
- the floor is identical for every channel, so the ranking is made by volume and price, never by preference;
- fits is a yes/no about your horizon, not about the channel's quality: a good channel outside its horizon is still the wrong next purchase;
- the first test is the cheapest two-arm test that fits — ties go to the faster one — and if nothing fits, the tool says so in plain words instead of shrinking the test.
The arithmetic is the published sample-size floor, not a forecast: compare with the same computation
in a terminal, npx marketing-mindset sample --base 0.03 --lift 0.2 (it prints per_arm
13914 and total 27828). Sibling tools: what one
honest test costs to finish, which funnel step can fund a
verdict, the honest test queue planner,
the A/B verdict calculator, the full list at
/tools/.
Written by Axel Freeman — marketing engineer. Free to use, no email required; the packages behind the method are public on the offer page ($900 Sprint, $1,900/month Engine, $2,900 Full Build).