SYS/RES
Calculators
Value of a test, acquisition, traffic and SEO — growth arithmetic.
Expected value of a test
Traffic, expected lift, average order value, win probability and implementation cost: the expected value of an experiment before you run it. Prioritise on money, not on gut feel.
SYS/IN
SYS/OUT
Expected value
€1,500
Under these assumptions the test is worth more than it costs — in money, not p-values.
- Extra conversions if the lift is realised
- 125
- Incremental revenue if the lift is realised
- €10,000
- Expected incremental revenue (× P(win))
- €3,000
How this is calculated
Formula
The expected value (EV) of a test is:
EV = T × CVR × λ × AOV × P(win) − C
where:
- T — traffic (sessions) exposed during the test window
- CVR — baseline conversion rate, as a decimal (2.5% → 0.025)
- λ — expected relative lift, as a decimal (10% → 0.10)
- AOV — average order value, in euros
- P(win) — probability the variant actually wins, as a decimal
- C — implementation cost, in euros
The product T × CVR × λ × AOV is incremental revenue if the lift is realised. Multiplying by P(win) gives the expectation; subtracting C gives net value.
This is not a p-value. A positive EV does not mean the test is significant; it means that, under these assumptions, the test is worth more than it costs.
Sources
Classic decision framing (expectation × stake − cost). Kohavi, Tang & Xu, Trustworthy Online Controlled Experiments (Cambridge University Press, 2020), on the opportunity cost of an experiment.
Limits
- P(win) is a judgement, not a Bayesian posterior computed here.
- Lift is assumed constant across all of T.
- The model ignores the opportunity cost of the split (some traffic stays on control).
- This is not a sample-size or duration calculation — see the CRO tools in this centre.