Channel Test Budget Calculator

A free tool for founders deciding how much to spend testing a new acquisition channel, and for how long, before calling it. You enter your target cost per customer, your weekly budget, and for up to six channels the cost per click (or CPM and click-through rate) and the share of clicks that become a lead. It returns the leads a fair read needs, what that costs, how many weeks it takes, a stop rule worked out in advance, and which channel to test first.

What it computes

• Cost per lead and per customer on your own numbers. Cost per click divided by the share of clicks that become a lead, then by the share of leads that become customers. If your own numbers already put a channel above your target cost per customer, it says so before you spend.
• The target cost per lead: your target cost per customer times your lead-to-customer rate. The test is judged on leads because customers arrive too slowly to judge a test on.
• The leads a fair read needs. Cost per lead is spend divided by leads, so its range is wider above than below. The tool finds the fewest leads whose exact Poisson range is within ±30% on both sides at 90% confidence: 46 leads. Precision and confidence can be changed.
• Spend and weeks to reach those leads at your weekly budget, with a two-week minimum so every day of the week is seen twice, and what a read on customers would take instead.
• A stop rule from the exact Poisson distribution, built to be checked once a week: if spend reaches a line with that many leads or fewer, stop. Each weekly check gets an equal share of the 10% error budget, so a channel that is really on target hits a line by mistake less than 10% of the time over the whole test.
• Which channel first: the testable channel that is cheapest to judge, or, when two are within 10% on that cost, the one with more room under your target.
• A test in progress: enter the spend and leads so far and it says stop, ahead of target, read complete, or keep going.

Why I size the test before it starts

In my experience, "we tried LinkedIn and it didn't work" usually means a few weeks, a few thousand dollars and a handful of leads. That is a test that stopped before it could tell, not a verdict. So I'd rather settle three things before any money moves: how many leads a fair read needs, what that will cost, and the spend at which to stop early. If the answer is more than you can afford, that is worth knowing too. A cheaper channel you can actually read is a better first test than the one everyone talks about.

Limits

• It is only as good as the rates you give it. Label guesses as guesses and re-run once the first week's numbers are in.
• A good cost per lead is not yet a good channel. Follow the leads to customers over one full sales cycle before you scale.
• The lead count, stop rule and progress checks use exact Poisson math and assume leads arrive at a steady rate. The stop rule assumes one check a week.
• The ±30% precision, 90% confidence, two-to-twelve-week window and 10% near-tie rule are AAJ's defaults, not industry benchmarks.

The same engine runs as an agent skill, channel-selection-and-testing. Once a channel works, split the budget with the Paid Media Budget Allocator. For your target cost per customer, use the Unit Economics Calculator. For the channels themselves, see the Digital Advertising Playbook.

Run this in your agent

The free AAJ agent skill Channel Selection & Testing does this job inside a coding agent, on your own data. Install it with npx skills add sarojkjha/aaj-marketing-skills --skill channel-selection-and-testing.

Using Claude or ChatGPT instead? Add the AAJ engines as a custom connector, with no install: https://aaj-engines-mcp.sarojkjha10.workers.dev/mcp