Most advice about A/B testing assumes you get thousands of visitors a day. A personal page gets dozens. That gap kills the standard method, and it is why most people who try to ab test homepage copy quit after a week: they run it by the book, eleven visitors show up, and the result says nothing.
There is a version that works at this scale, and it is closer to asking people a question than to solving a statistics problem. Below: what is actually worth testing on a one-page site, the smallest homepage experiment that returns an answer, and how to read the result without talking yourself into it.
## Three things worth testing on a one-page site
Three things are cheap to swap and move the needle: the intro line, the order of sections, and the primary action.
The intro line moves more than anything else on the page. It is the first thing most visitors read and the only thing they repeat to someone else. Changing it costs five minutes and touches nothing else.
Section order comes second. Portfolio above the bio and bio above the portfolio are two different pages, and switching between them is a drag.
The primary action is third. "See my work" and "Email me about a project" invite two different visitors, and you will not know which one is arriving until you try both.
What you cannot test at this volume: colour, typeface, grid. Those need traffic you do not have. Anyone reporting that their accent colour lifted conversions by 12 percent on a personal page is guessing.
## The smallest experiment that returns an answer
| Variable | Version A | Version B | Run each | What counts as a result |
|---|---|---|---|---|
| Intro line | "I redesign signup flows for small SaaS teams" | "Product designer, eight years, mostly B2B" | 2 weeks | Someone repeats the line back to you |
| Section order | Portfolio above bio | Bio above portfolio | 2 weeks | Share who scroll past the portfolio grid |
| Primary action | "See my work" | "Email me about a project" | 2 weeks | Enquiries that mention the page |
Two weeks each is the floor, not a suggestion. In a single week the difference between Tuesday and Saturday is larger than the difference between your two versions.
Test one thing at a time. Change the intro and the order together and you get two numbers that both improved slightly, with no way to tell which one to keep.
## Running it without an A/B tool
Run them in sequence, not at the same time. With a dozen visitors a day you cannot split the traffic, but you can compare two clean fortnights, as long as nothing else changed in between.
Sequential testing has a cost: the traffic sources in fortnight one and fortnight two may differ. Hold that in mind rather than either ignoring it entirely or treating it as the finding.
Write the swap date in your changelog. That one line saves you in two months, because by then you will not remember which week you changed it.
## How to read the result without fooling yourself
Decide what counts as winning before you start. Decide after, and you will always find an angle that supports whichever version you prefer, which is almost always the one you wrote second.
Use behaviour, not feeling. "Someone repeated my line back to me" can be counted. "This version looks more professional" cannot.
With a small sample, the gap has to be obvious before it means anything. Three enquiries versus five is not a 67 percent lift. It is two enquiries.
No difference is also a result. It means neither version sits on the path that decides what a visitor does, and the thing to change next is usually the images rather than the words.
## FAQ
- Is this worth doing with only twenty visitors a day? Yes, if you accept a coarser answer. You will learn that one version is clearly better. You will not learn that it is 8 percent better.
- How long should one version run? Two weeks. One is too short and four is too long, because by week three something else has usually changed on the page.
- Do I need analytics installed? You need to know at least how many people scroll past the portfolio. Pro's visitor analytics covers that and is less hassle than adding a third-party script.
- What if nothing changes? Switch variables. If the intro, the order, and the action all move nothing, the next thing to look at is the work itself rather than the words around it.
## Start one today
Open your page, rewrite the intro line into the more specific version, note today's date in your changelog, then leave it alone for two weeks. The whole thing takes five minutes, and the hard part is not touching it in between.
On dynamic-profile.shop changing one line does not rebuild the page, which is what makes leaving it alone for two weeks realistic. Start from the [homepage](/). Set your baseline first with [the five-second test](/blog/five-second-test-first-impression-homepage). For which number to watch, [the visitor analytics piece](/blog/visitor-analytics-101-reading-data) covers reading the data, and if you cannot write two versions, [cutting your homepage copy in half](/blog/cut-homepage-copy-by-half) has the shorter checklist.