With little traffic, test a bold price change on half of your new visitors. Small tweaks can take years to read.
- At 4,000 visitors a month, a 50% difference shows in about 6 weeks. A 20% one takes about 6 months.
- 50% of visitors on the test price is about three times faster than 10%.
- Below 1,000 pricing page visitors a month, talk to customers instead.
What sets the duration
A price test needs enough sales on each price to tell real differences from luck. Three things decide how fast you get them:
- How many visitors reach your pricing page and buy.
- How big the difference between the two prices is.
- How many visitors see the test price, since the smaller side sets the pace.
The size of the difference matters most. Halving it makes the test take about four times longer. Time to plan for at 3% of visitors buying, half of them on the test price:
| Visitors / month | 20% difference | 30% | 50% |
|---|---|---|---|
| 2,000 | 12 months | 6 months | 3 months |
| 4,000 | 6 months | 3 months | 6 weeks |
| 10,000 | 2 months | 5 weeks | 2 weeks |
2,000
- 20% difference
- 12 months
- 30%
- 6 months
- 50%
- 3 months
4,000
- 20% difference
- 6 months
- 30%
- 3 months
- 50%
- 6 weeks
10,000
- 20% difference
- 2 months
- 30%
- 5 weeks
- 50%
- 2 weeks
Your own numbers: test duration calculator.
Five levers that speed it up
1. Test a bold change
$29 to $39 is a 34% higher price. Even if a few fewer people buy, the gap in revenue per visitor is big enough to see. $29 to $32 rarely is. If you’re not sure where to go, go further than feels comfortable: a clear loss is a useful answer too.
2. Put 50% of new visitors on the test price
The verdict waits for the smaller side. At 10%, that side fills five times slower than at 50%. Start at 10% for a day if you want to watch it work, then raise it.
3. Test the plan people actually buy
One test, on the plan with most sales. A test on a plan that sells twice a month will never finish.
4. Measure revenue per visitor
It uses every visitor and every sale, and it answers the real question: which price earns more. Waiting for a difference in conversion rate alone would take even longer.
5. Don’t peek and stop
Small tests swing a lot in the first weeks. Stopping when a lead looks good is how founders keep the wrong price.[1] PricingFast holds the verdict until 95% confidence and 30 sales.
Worked example
RunStack, a visual LLM workflow builder, gets about 1,500 visitors a month on its pricing page. Its founder tests the Starter plan at €39 instead of €29, on half of new visitors.
The test price earns 29% more per visitor.
| Price | Visitors | Sales | Per visitor |
|---|---|---|---|
| €29/moToday’s price | 2,900 | 96 | €0.96 |
| €39/moTest price | 2,900 | 92 | €1.24 |
Verdict reached. A €10 step was big enough to read on 1,500 visitors a month. A €3 step would still be running next year.
What slows a small test down
- Testing three or four prices at once. Each one splits your traffic again.
- Testing several plans at the same time on the same visitors.
- Counting all site visitors. Only people who reach the pricing page see a price.
- Running a test across a big launch or a sale, which changes who visits.
Below 1,000 visitors a month
At this size, even a bold test can take most of a year. For comparison, self-serve products that convert 3–5% of free users to paid are considered good.[2] A few hundred visitors bring a handful of sales a month.
- Ask your last 10 customers what they would have paid, and what almost stopped them.
- Raise the price for new customers only and watch revenue per visitor month over month. It’s weaker evidence than a test, but it’s evidence.
- Come back to testing once traffic grows: the calculator tells you when it’s worth it.
Questions
Can I A/B test pricing with 1,000 visitors a month?
Only for very large differences, and it will take many months. Below about 1,000 pricing page visitors a month, talking to customers and raising the price for new customers is usually a better use of time than a formal test.
What's the fastest way to get a pricing test result?
Put 50% of new visitors on the test price and test a bold change. At 4,000 visitors a month and 3% buying, a 50% difference in revenue per visitor can be read in about 6 weeks; a 20% difference takes about 6 months.
Should I lower my confidence threshold to finish sooner?
PricingFast keeps 95% for its verdict, but you can end a test at any time and keep either price. The dashboard shows how sure it is, so you decide knowing the risk instead of reading noise as a result.
Is a 50% split risky?
Less than it sounds. Only new visitors see the test price, existing customers keep theirs, and pausing puts everyone back on your usual price on their next page load.
Sources
- [1]How Not To Run an A/B Test (Evan Miller), evanmiller.org
- [2]What is a good free-to-paid conversion rate?, Lenny’s Newsletter, 2023