Exactly. I deal with math and numbers all day at work, and there's a million ways you can slice and dice data. I had a customer the other day tell me: "I don't think the data looks right on this chart.". I told them the data was EXACTLY right, given how our software works, the math transformations they applied to their underlying data, and the shape of the underlying data itself.

Same thing here. You have to make sure you are looking at the right metrics to make the right decisions.

For example:

• Don't use a single mailing to determine anything. Too small a sample size makes variance too wide. UNLESS the results are so obvious that it's a no-brainer.
• Run the same mailing more than once, then take an average.
• And if you see spikes and dips, ask why. Look at the context.

I should post more about data interpretation. I love this stuff.