Low-traffic sites

You do not need more traffic. You need a bigger question.

The visitors a test needs fall with the square of the effect it is looking for. Test one word and the effect is tiny, so the traffic required is enormous. Test an argument told across the whole page, and it collapses.

What it would test first

On a small site the engine picks fewer, larger questions rather than a queue of small ones.

Example · a one-person consultancy

Theme 01 · Who it is for

Naming the specific reader will do more than describing the service, because a small audience converts on recognition rather than on breadth.

For founders who cannot afford a bad hire.
You are two engineers and a spreadsheet. Start here.

Theme 02 · What happens next

Removing the uncertainty about the first step will do more than another proof point, because hesitation on a small site is usually about process rather than trust.

A thirty minute call. No deck, no pitch.
Tell us the problem. We will say if we can help.

Two arguments for the same thing, each written more than one way. What gets tested is which argument lands, not which sentence reads better.

Why it works down here

Nothing about this is a smaller version of the enterprise product. It is a different trade.

One question at a time: Traffic is not split across a queue of concurrent tests. One question gets all of it, which is the difference between an answer and a shrug.
It waits: A winner is called on the strength of the evidence, not on a schedule. On a quiet site that takes longer, and the engine is content to wait rather than promote something it cannot support.
A tie still counts: Most tests on a small site come back with no difference. That gets filed against the idea rather than thrown away, so your traffic never buys the same answer twice.
What you learn keeps: A belief about what your audience responds to survives a redesign, a new season and a change of theme. That is why patience costs you less here than it would with a tool testing layouts.
A bigger question needs less traffic: Not proportionally less. Squared. The visitors a test needs fall with the square of the effect it is looking for, so an idea that moves conversions twice as much needs a quarter of the traffic, not half. On a page converting at ten percent, hunting a ten percent lift takes about 14,000 visitors per version; hunting a twenty percent lift takes about 3,500. Testing one word is the expensive way to test, because the effect is small and small effects are what traffic gets spent on.
One idea, told across the whole path: A hypothesis is not confined to one element. The same argument can be made in the headline, the section under it and the page the button leads to, which moves a larger number than any of them would alone. That is the lever a small site has: not more visitors, but a bigger question asked of the ones it gets.

Try it on your own site

Paste a URL and A/Bee hands back its first hypotheses for that page. Free, and no signup.

See what it would test

What is a belief, exactly?