Is that weekday pattern real?

One synthetic advertiser’s cost per sale. You set the weekly rhythm that is really there and how much history you get to see it with, and the method tries to recover it. Set the real rhythm to zero and turn the drift up: the pattern it reports is then entirely manufactured. Synthetic data throughout.

Weeks of daily figures: 6

The weekend effect that is really there: +17%

Account drifting: not at all

what the method recovered the rhythm that is really there

Show the numbers as a table

Day	Recovered	Really	Out by	Observations

Self-check: running…