Why order matters
Every site has more test ideas than it has traffic or time. A small site has even less of both. If you spend your limited visitors on a minor change, such as a slightly different button shade, you may wait weeks and learn nothing useful. Deciding what to test first is therefore as important as how to run the test.
The goal is simple: spend effort where a change could plausibly matter to the business and where you can make it without a large project.
Start with where visitors are
You can only learn from places that have visitors. Look at your analytics to find the pages that get the most traffic and the ones that sit just before a key action, such as a contact form, a booking or a checkout. A page with plenty of visitors and a clear goal gives you the best chance of seeing a result in a reasonable time. Pages with very few visits may not support a formal test at all; the guide to how much traffic you need explains why.
It also helps to look at the funnel drop-off. If most people leave on one particular step, a change there has more room to help than a change on a step that already works.
Gather clues before guessing
Ideas are better when they come from evidence. Useful sources include:
- Customer feedback and questions that arrive by email, chat or phone.
- Analytics showing high exit rates or low form completion.
- Watching a few real people try to use your site, even informally.
- Reviewing your page with a checklist, for instance for clarity of the main message.
If many customers ask the same question, the answer may belong in a headline or near a call to action.
A simple scoring method
Give each idea two scores from 1 to 5.
- Impact: how much could this change plausibly move the main goal? 1 is barely noticeable, 5 is large.
- Effort: how hard is it to build, review and launch? 1 is a quick edit, 5 is a big project.
Then divide impact by effort. Higher is better. Here is an illustration with invented scores:
| Idea | Impact | Effort | Score (impact / effort) |
|---|---|---|---|
| Rewrite the main headline to state the offer clearly | 4 | 1 | 4.0 |
| Shorten the contact form | 3 | 2 | 1.5 |
| Redesign the whole page | 5 | 5 | 1.0 |
| Change the footer link colour | 1 | 1 | 1.0 |
On these scores the headline test comes first, the form second, and the redesign ties with a trivial change. The scores are judgement calls, so treat the ranking as a conversation starter, not a calculation to defend. If two people disagree on a score, that is useful information.
Some teams add a third score for confidence: how strong is the evidence behind the idea? An idea backed by repeated customer comments deserves more weight than a hunch.
Good first candidates
These elements are often worth considering, although what works varies by site:
- The headline and opening message. It is what visitors read first, and clear wording is cheap to test. See headline writing for ideas on what to vary.
- The call to action. Wording, placement and how many competing actions are on the page.
- Form length and fields. Each extra field can add friction.
- Offer and framing. What the visitor gets, stated plainly, and how risk is handled.
- Page layout and clarity on a key landing page, especially on phones.
- Email subject lines, if you send regular campaigns and have a decent list.
Bigger, bolder differences are easier to detect on small traffic than subtle tweaks, so lean towards changes that visitors will genuinely notice.
What to leave for later
- Tiny cosmetic changes with little expected effect.
- Pages with very few visitors.
- Multi-element experiments that need lots of traffic.
- Ideas you cannot state as a clear test hypothesis.
Turn the winner of the ranking into a test
Take the top idea and write the hypothesis: what you will change, what you expect, why, and which single metric will decide. Then plan the sample size and duration, run the test without peeking, and record the result in your testing log. Move to the next idea on the list, and re-score the list now and then as you learn.
Keep expectations realistic
Prioritisation improves your odds, but it does not guarantee a win. Many tests show no meaningful difference, and results depend on your audience and offer. If testing is impractical at your traffic level, you can still use the same ranking to decide which improvements to make directly, based on feedback and good user experience practice. Either way, you will be putting effort where it is most likely to count.