A pricing test needs enough sales, not enough days: at 10,000 visitors a month, plan for about 2 months.
- The verdict needs 95% confidence and at least 30 sales.
- More visitors on the test price, or a bolder price change, makes it faster.
- Put in your own numbers below.
New visitors per month. Your analytics has it.
Pricing page visitors who become paying customers.
Share on the test price
You pick this when you press Launch.
Smallest difference worth finding
In revenue per visitor. Bigger is faster to see.
Plan for 2 months.
That’s about 24,389 visitors, 12,195 of them on the test price, and 732 sales. If the difference is real, there’s an even chance of a verdict by 5 weeks. Set by the size of the difference you want to see.
Plan for, with other choices
| Share | 10% | 20% | 30% | 50% |
|---|---|---|---|---|
| 10% | ||||
| 25% | ||||
| 50% |
Uses PricingFast’s verdict rules: revenue per visitor, 95% confidence, at least 30 sales. Planned so a real difference of that size is found 4 times out of 5.
How it’s calculated
The calculator uses the same rules as the PricingFast dashboard. A test ends with a winner when one price has at least a 95% chance of earning more per visitor, there are at least 30 sales, and each price has had at least 100 visitors.
It plans the test so that a real difference of the size you pick is found 4 times out of 5 (80% power, in statistics terms). The “even chance” line is the point where half of such tests would already have a verdict. Real tests land somewhere between the two.
- Revenue per visitor is the metric, not conversion rate: a higher price that sells a little less can still win.
- Both prices are assumed to convert at a similar rate. If the test price converts much less, the real test takes somewhat longer.
- The price amount doesn’t matter. Only the share of visitors who buy and the size of the difference do.
Durations at a glance
Time to plan for at 3% of visitors buying and half of them on the test price. Use the calculator for other splits.
| Visitors / month | 10% difference | 20% | 30% | 50% |
|---|---|---|---|---|
| 2,000 | over 3 years | 12 months | 6 months | 3 months |
| 4,000 | 22 months | 6 months | 3 months | 6 weeks |
| 10,000 | 9 months | 2 months | 5 weeks | 2 weeks |
2,000
- 10% difference
- over 3 years
- 20%
- 12 months
- 30%
- 6 months
- 50%
- 3 months
4,000
- 10% difference
- 22 months
- 20%
- 6 months
- 30%
- 3 months
- 50%
- 6 weeks
10,000
- 10% difference
- 9 months
- 20%
- 2 months
- 30%
- 5 weeks
- 50%
- 2 weeks
A 10% difference is hard to see below tens of thousands of visitors. That’s why small SaaS should test bold price changes: here’s how.
If the answer is too long
- Put 50% of new visitors on the test price instead of 10%. It roughly triples the speed.
- Test a bigger price change. Going from $29 to $39 is easier to read than $29 to $32.
- Test the plan most people buy, not the one that sells twice a month.
- Count the visitors who reach your pricing page, not your whole site: that’s who sees a price.
Questions
How long should a pricing test run?
Until it has enough sales to be sure, not for a fixed number of days. PricingFast calls a winner when one price has at least a 95% chance of earning more per visitor and the test has at least 30 sales. With 10,000 pricing page visitors a month, 3% buying and half of them on the test price, plan for about 2 months to find a 20% difference.
How many visitors do I need for a pricing A/B test?
It depends on how many visitors buy and how big a difference you want to find. At 3% conversion with a 50/50 split, finding a 20% difference in revenue per visitor takes about 24,000 visitors; a 50% difference takes about 5,200.
Why does the share on the test price matter so much?
The verdict is limited by the smaller side. With 10% of visitors on the test price, that side fills five times slower than at 50%, so the same test takes roughly three times longer overall.
Does the price itself change the duration?
Barely. Revenue per visitor scales with the price, so its relative noise depends on how many visitors buy, not on whether you charge $9 or $900. The calculator assumes both prices convert at a similar rate.
Can I stop a test early if it looks like a winner?
In PricingFast the verdict only appears once both rules are met: 95% confidence and 30 sales. Until then the dashboard shows which price is leaning ahead but asks you to wait. Stopping on an early lead is the most common way to keep the wrong price.