Knowing whether it helped.
THE PROBLEM WITH SMALL SITES
Not enough traffic for formal testing.
WHAT TO DO INSTEAD
Change one thing, compare periods.
Accept that it is imprecise.
WHAT MAKES COMPARISON UNRELIABLE
Seasonality A campaign running in one period Different traffic sources
WHAT TO CONTROL FOR
Compare like periods where you can.
FOR SITES WITH ENOUGH TRAFFIC
Split testing, where half see each version.
That removes timing effects.
WHAT TO TEST
One element: headline, action, form length, price display.
HOW LONG
Until the difference is clearly larger than normal variation.
WHAT TO BE HONEST ABOUT
Small differences on small numbers mean nothing.
WHAT NOT TO DO
Declare a winner on twenty visits Test trivial things while large problems remain
