Our business rents products to customers using a variety of payment plans.. ranging from 3-month (high \$ per month) to a continuous rental (low \$ per month). I've created a spreadsheet that shows how much revenue is generated by month by orders that were approved on a certain month.
More specifically... across the header row, I have different months from the past 5 years. The first cell in each column is the revenue generated in the month the order was approved. The second cell in each column shows the revenue generated the second month. Third cell shows revenue from the third month... etc.
Due to order fall-out and the varying rental plans used, revenue from orders approved in a specific month decomposes logarithmically. The next cell in each column would be the revenue for the next month not included in the model. If I can predict the next value in each column, I should be able to draw a forecast of next months revenue.
I'm looking for advice on how to fit a model to a month where we have adequate history (at least 15 months ago -- providing 15 data points), and use that fit to model future months in columns which we don't have many cells (e.g., last month). I've tried taking the natural log of each cell and fitting a line to each column using for-loops. However, I figured I could get a better representation if I used a line fit to a month with an adequate number of values to predict new values in columns with fewer values.
I will be using R for this implementation. I'd appreciate any advice you can offer -- whether it be for my current idea or if you have a different approach.