Lapizuli - 4:29 pm on Apr 9, 2010 (gmt 0)
It's a matter of statistical accuracy. Assuming that significantly greater revenue means more content and longevity (a fair assumption in most cases), Websites bringing in $1000 a day will tend to have a lower standard deviation - i.e., variation from the average - than those bringing in $20 a day. An extreme change in the $1000 income will have more statistical significance.
Statistical significance is an indicator of how useful data is for noticing trends.
A sufficient sample size is important if you're going to point to a trend that is "real" - i.e., predictive. The greater the sample size, the more your results approach consistency, all other things being equal. Nicely consistent data based on a large sample size suggests that that all the random factors that go into bringing about variations in earnings average out. Those fluctuations may be important on one level, but they are "consistently variable," so to speak.
So if, say, for three years you've been bringing in $1000 a day, plus or minus $50, you have some useful data to work with. When you're suddenly earning $800 daily for a week, you're more likely to be noticing evidence of a real change in the system than a website that earns only $20 per day seeing a $10 loss, even though you only noticed a 20% drop and they noticed a $50% drop. The smaller website's variations can be looked at as a statistical artifact of having a low sample size.
People are designed to notice patterns, so we tend to see meaning in numbers that are often not statistically meaningful. Meaningful in other ways :) but not statistically.
And all that said...statistical significance can be seen as a relative thing, too. I step on one bug, which has no statistically significant affect on the bug population, but was very significant for that bug.
When Google AdSense changes its earnings algorithm, the data may show overall no notable change in revenue, and so in Google's eyes, what they did wasn't a big deal. But some individual publishers, large and small, will be wiped out and others will get rich. For those small businesses, the change caused statistically significant variations.