Should you audit, fix or A/B test?
Enter the visitors and enquiries for one page. The tool estimates how many visitors a two-version test would need and how many weeks that would take.
There is no magic traffic number that makes a site ready for testing. The answer depends on your conversions, the change you need to detect and how long you can wait. Sometimes the honest answer is that nobody can tell yet, and the tool says so.
Decision aid
Audit, fixes or an A/B test?
An A/B test only helps if it can reach an answer in a sensible time. That depends on how many conversions you get, how big a change you need to detect and how long you can wait. Monthly traffic on its own does not decide it.
We cannot tell yet
Enter the visitors and enquiries for the page you have in mind. If you do not know them, that is the answer for now: measure first.
A conversion audit starts with the tracking, so you leave with numbers you can rely on.
Ask about a conversion auditThe method
The estimate is the standard sample-size calculation for comparing two conversion rates: a two-sided test at 5% significance with 80% power, and visitors split equally between two versions. Weeks are rounded up, because a test should run in whole weeks to cover weekdays and weekends alike.
It is a planning estimate. It does not allow for seasonality, several variants, or checking the result early, all of which make a test take longer or mean less.
Worked example with invented figures
A page with 2,000 visitors and 40 enquiries a month converts at 2%. To detect a 20% relative improvement, each version needs about 21,100 visitors, so 42,200 in all. At 2,000 a month that is around 92 weeks. A test is the wrong tool for that page today.
The same page with 60,000 visitors and 1,200 enquiries a month would reach the answer in about four weeks.
What the four answers mean
- We cannot tell yet
- There are no figures, or no recorded conversions. Measure first: test the form, check the enquiries arrive and check they are counted once.
- Check the tracking first
- The sums work, but you have not confirmed the figures. A test built on miscounted enquiries gives a confident wrong answer.
- Research-led fixes first
- A test would take longer than you are willing to wait. Improve the page with evidence that does not need a split test, and measure before and after with the limits stated.
- A controlled test is plausible
- The numbers could support a test. It still needs a clear hypothesis, a sample size fixed in advance and a rule for when to stop.
To see what an improvement would be worth, use the conversion scenario calculator. For the service behind the tool, see conversion rate optimisation, conversion rate optimisation in Manchester or what it costs.