How to minimize variability in order to design a valid pricing experiment?

#1
Good morning,

I want to run a pricing experiment to see the change in the number of purchases (i.e.demand) in response to a 10% increase in price in a specific market for my product. I selected 2 markets with very similar characteristics, where one will get a treatment effect and the other will serve as a control (no effect) for a duration of the experiment (i.e. one week).

However, when I looked at the historic data, I can see that the number of purchases has a lot of variability week by week for both markets.

In particular, for each market, the number of purchases fluctuates by about 10% up and down from week to week. The trend seems completely random and is not the same for both markets. Given the nature of the business there does not seem to be a specific factor that would cause so much variation from week to week for either one of them.

How can I control for this variation to make the experiment valid? My plan was to run regressions of demand on price, but the problem is that I think my standard errors will be too high to discern the effect of the price coefficient. What should I do?

Thank you!