What's going on here?

Each day of the week (Mon, Tue, Wed...) occurs a different number of times in different months, depending on how the week boundaries happen to line up with month boundaries. And this alignment changes each year. So if some days of the week tend to be higher-performing than others (such as restaurant sales with more customers on the weekends), monthly summaries become unreliable simply because of differences in the number of higher-performing week days that happen to fall in each month.

In the example above, a constant rate of growth (2% per year) is easily seen in weekly summaries but masked in monthly summaries. The monthly noise may cause decision makers to erroneously conclude that they were responsible for the positive or negative growth.