Unit economics and test planning
CAC, LTV and revenue calculators tuned to the vertical you actually run, plus the sample size an A/B test needs before it can answer anything. Defaults are KZ-realistic; the math is the kind your CFO would actually accept.
Acquisition
Spend per new userLoans
main LTV driver; counted once per lifetime, not monthlyDeposits
silent but powerful LTV multiplierTransfers
small fee × frequencyPayments (card, QR)
interchange × volumeFAQ
How do I calculate CAC?
CAC (Customer Acquisition Cost) = total sales and marketing spend in a period divided by new customers acquired in the same period. Include ad spend, marketing salaries, agency fees and tooling. Exclude organic and existing-customer revenue.
How do I calculate LTV?
LTV (Lifetime Value) = average revenue per customer per period × gross margin × expected customer lifetime. For subscription businesses use ARPU × gross margin ÷ churn rate. For e-commerce use AOV × purchase frequency × the area under the retention curve.
What is a healthy LTV/CAC ratio?
Rule of thumb: 3:1 or better for SaaS and subscription. 1.5–2:1 can work for high-volume e-commerce with fast payback. Below 1:1 you are losing money on every acquisition. Always pair it with payback period — a 5:1 LTV/CAC with a 36-month payback is a cash crisis.
How does payback period interact with CAC?
Payback = CAC ÷ monthly gross profit per customer. Shorter payback means less working capital tied up in growth and lower funding risk. Fintech apps typically target 6–12 months, e-commerce often runs 1–3, SaaS 12–24.
What is a minimum detectable effect?
The smallest difference the test is designed to catch, expressed here relative to the baseline: a 10% MDE on a 4% conversion rate means the test can resolve a move to 4.4%. A smaller effect can still be real — it just won't be distinguishable from noise at the sample you planned.
Why does an A/B test need so much traffic?
Required sample scales with the inverse square of the effect. Cutting the effect you want to detect in half multiplies the sample by four, and that is before traffic is split across more than two arms.
Can I stop the test as soon as it turns significant?
Not on these numbers. The formula assumes a single test at a fixed sample size. Peeking daily and stopping at the first significant result can push the real false-positive rate to 20–30%. Either commit to the planned sample or use a sequential method designed for continuous monitoring.
How do I handle three or four variants?
The sample is per arm, so total traffic and duration scale with the number of arms. Each extra arm is also another comparison against control, which raises the chance of at least one false winner: divide α by the number of comparisons (Bonferroni) if the family-wise error rate has to stay at 5%.
The formulas are fine. The inputs usually aren't.
CAC, LTV and payback are arithmetic. The hard part is where the numbers come from: ad spend sits in six platforms, revenue sits in the CRM, refunds sit nowhere, and nothing shares a join key. A model built on numbers you can't reconcile is a confident wrong answer.
Field notes: The end of rented audiences: why businesses are building their own DWHs and collecting First-Party Data
The document we fill in before anyone opens a container: naming conventions, identity model, event and property tables with worked examples, a destination map, and the QA checklist that catches double-counted revenue.