I am doing time series forecasting for estimating monthly sales of certain consumer goods SKU.

I have last 3 years of actual sales data, wherein first 2.5 years of data is monthly sales data, but last 6 months data daily sales data is available.

I can certainly add the daily data to get monthly data, but considering that total data points are then only 36 (for 3 months) and it may be slightly challenging to get various models like ARIMA, Exponential models.

Is there a way to use the daily last 6 months of data to improve the accuracy of the model?


I would do two separate analyses:

  • By day for six months
  • By month for the last 36 months (with the most recent 6 aggregated)

The 36 month analysis will be important for catching annual trends you wouldn't be able to see with the daily data, but you'll have a lot more fine-grain with the daily analysis to catch quarterly, monthly, and weekly cycles.

  • $\begingroup$ If this helped, don't forget to mark it answered :) $\endgroup$ – CalZ Oct 5 '17 at 20:31
  • $\begingroup$ How to do mark it answered CalZ? $\endgroup$ – AKM Oct 7 '17 at 1:37
  • $\begingroup$ I believe there should be a check mark next to the vote totals to the left of your post. $\endgroup$ – CalZ Oct 7 '17 at 13:35
  • $\begingroup$ Done CalZ.. Thanks.. Am new to Stackexchange.. Shall ensure I do that for any query that gets clarified here.. $\endgroup$ – AKM Oct 9 '17 at 8:09

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.