A/B testing ideas for your website

The real cost of testing wrong things

Your testing budget is finite

Your website gets ten thousand visitors per month. One test takes two weeks and uses up two thousand visitors to get statistical significance. You can run six tests per month maximum.

Six tests per month is seventy-two tests per year. If all are winners, seventy-two improvements compound to massive growth. If most are low-impact tests you run out of tests before tackling high-impact problems.

Test the wrong things and you waste your testing budget on button color when your biggest problem is form abandonment.

Impact framework: what actually moves the needle

High-impact tests

High-impact tests. Five to thirty percent conversion improvement possible. Headline. Form length. Offer. Call to action. Price. These change business metrics noticeably. These go first.

Medium-impact tests

Medium-impact tests. One to five percent improvement possible. Button text. Copy. Secondary messages. Page flow. Worth testing but after high-impact.

Low-impact tests

Low-impact tests. Less than one percent improvement possible. Button color. Font. Spacing. Worth testing eventually but last.

The compounding difference

Different tests have vastly different potential impact. Changing your offer from free trial to money-back guarantee might improve conversion thirty percent. Changing button color from blue to green might improve by zero point five percent.

One high-impact test equals sixty low-impact tests in terms of business value. Prioritize ruthlessly.

High traffic and low conversion: your gold mine

Not all traffic is equal

Page one. Thousand visitors per month. Point five percent conversion. Five conversions. Bad but not critical.

Page two. Thousand visitors per month. Twenty percent conversion. Two hundred conversions. Great.

Page three. Three thousand visitors per month. One percent conversion. Thirty conversions. Worst conversion rate but highest traffic.

Test page three first

Test page three first. Even small improvement scales. One percent improvement on page three is three extra conversions per month. One percent improvement on page one is zero point zero five extra conversions. Page three improvement is sixty times more valuable.

Find pages with high traffic and low conversion. Those are your gold mines.

Build your testing backlog from business problems

Problem-driven testing

Do not ask what to test. Ask what business problem hurts most.

Problem one: checkout abandonment

Problem: checkout abandonment is forty percent. Customers get halfway through and leave. Test solution: simplify checkout. Single page instead of three pages. High impact. Goes first.

Problem two: low form submissions

Problem: form submissions are low. Even though traffic is high. Test solution: reduce form fields. Five fields instead of ten. Medium impact. Goes second.

Problem three: email decline

Problem: email open rate is declining. Test solution: different subject line format. Low-medium impact. Goes third.

Problem four: homepage scroll

Problem: homepage visitors do not scroll down. Test solution: move key value prop up. Low impact. Goes last.

Each problem gets a test. Backlog is problem-driven not random.

Priority scoring: PIE framework

P: Potential impact

Rate one to ten. If this test wins, how much will conversion improve. Offer test scores high. Button color scores low.

I: Importance of element

Rate one to ten. How important is this page or element. Checkout is more important than footer. Homepage headline more important than sidebar text.

E: Ease of testing

Rate one to ten. How easy is this to test. Button color test is easy. Checkout redesign is hard.

Calculate the score

Score. Multiply P times I times E. Highest score is highest priority.

Example. Checkout test. P equals nine (big impact), I equals ten (critical page), E equals five (complex). Score four hundred fifty.

Example. Button color test. P equals one (tiny impact), I equals eight (homepage), E equals ten (easy). Score eighty.

Run checkout first. Button color much later.

What most businesses test (wrong)

Aesthetics are easy but low-impact

Most businesses test aesthetics. Button color. Font size. Spacing. These are easy to test so people test them. They are also low impact so people waste testing budget.

Mechanics are hard but high-impact

Smart businesses test mechanics. How form works. What offer is. How checkout flows. These are harder to test so less competition. They are also high impact so better ROI.

Test mechanics before aesthetics.

When to skip a test entirely

Traffic too low

Test needs minimum sample size. If you get fifty visitors per month, do not bother. You will not get results for a year.

Change too subtle

Visitors will not notice. Testing something that makes no visible difference wastes time.

Unclear what you are measuring

What are you measuring. If you do not know what matters do not test.

No implementation plan

If treatment wins will you actually implement. If not skip test.

Already know the answer

If everyone agrees this will work stop guessing and implement.

Frequently asked questions