# Would a time series model be appropriate for this problem statement?

My problem statement is as follows:

I have thousands of customers and have data for each of the customers for the past 30 months. What I want to do is predict the sales of each of these customers for the next six months.

My current set up:

I built several lag variables and created response variables that are essentially a 6 month lookforward from the given month. See table below for an example of one customer (numbers represent the data for a specific month):

customer    Predictors (62 Features)    Response
111111      13                          19
111111      14                          20
111111      15                          21
111111      16                          22
111111      17                          23
111111      18                          24


I'm not getting the greatest results when I try this so I'm looking for some alternatives. Any help would be much appreciated

EDIT:

Just to clarify, for each month, each customer will have approx 62 features.

• You tagged your post as multivariate, but your question seems to indicate that you have only one variable per customer. – Skander H. Apr 9 '18 at 23:05
• So there are in fact multiple features fort each customer. I've added edits that I hope clarify this. – madsthaks Apr 9 '18 at 23:12