Four calculators that run in your browser
Conversion math that states its limits.
Test results, sample sizes, funnels and break-even, with the assumptions shown next to every number. No server, no tracking, no stored data.
Set up and embed
Host this standalone file on a site you control and embed it with an iframe. Do not paste the whole document into another page, because it brings its own styles and script. The embed needs scripts allowed (a sandbox attribute must include allow-scripts, and allow-downloads if you want the download button to work), and the frame height may need adjusting on narrow screens because results grow with content. Money inputs share one currency; nothing is converted.
<iframe src="https://your-site.example/conversion-kit.html?tool=ab" title="A/B result calculator" width="100%" height="900" style="border:0"></iframe>
?tool= accepts ab, size, funnel or roi. &theme=light or dark forces a theme. The functions are exposed for testing as ConversionKit.abTest, sampleSize, funnel and campaign; each returns {error} for invalid input.
Methodology and limits
A/B result calculator
Two-sided pooled two-proportion z-test (NIST handbook). Pooled rate p = (xA+xB)/(nA+nB); z = (pB-pA) / sqrt(p(1-p)(1/nA+1/nB)); p-value = 2(1-Φ(|z|)), bounded to 0 to 1. The interval is an approximate unpooled interval, and verdicts use the unrounded p-value, so the two can disagree near the boundary.
The p-value is the probability of data at least this extreme if both versions had the same true rate and the assumptions hold. It is not a probability about either version, and 1-p is not the chance that B wins.
Assumptions: independent randomized units, one prespecified metric, and a fixed sample size read once. There is no correction for sequential testing, so repeated peeking inflates false positives. When any pooled expected count is below 5, or the pooled rate is 0% or 100%, the z statistic, p-value and interval are shown as unavailable and no verdict is given.
Sample-size planner
Equal arms, two-sided: n = (zα/2·sqrt(2p̄(1-p̄)) + zβ·sqrt(p1(1-p1)+p2(1-p2)))² / (p2-p1)², p̄ = (p1+p2)/2. Per-arm n is rounded up; the total is twice that. Relative effect: p2 = p1(1+effect). Absolute effect: p2 = p1+effect in percentage points. It plans for an increase. Duration assumes steady traffic split evenly; run whole weeks.
Funnel and campaign economics
Counts are expected values from your rates, not forecasts. Gross revenue is orders times average order value, gross profit is revenue times margin, and net contribution is gross profit minus the ad spend you entered, with all other costs excluded. Break-even conversion is spend divided by clicks times profit per order; above 100% means spend cannot be recovered. Fixed costs, refunds, tax and repeat purchases are not modelled. Nothing here guarantees a result.