You could use a x weeks moving average, to hide the weekly spikes that distort your analysis. Maybe with 10 weeks moving averages, comparing to the 10 weeks average of the same weeks last year, you can calculate a much more reliable sales number that shows tendencies but hides spikes. Take a look at the green line, it shows the tendency, but makes it much softer. If you want to follow the changes faster, you could use a smaller number of weeks, as 5.







