Forecast based on monthly figures - sql

Question

I receive a monthly list of unpresented cheques with payment date. Those reached certain age in Month1 (say, 90 days to date) are recognized as overdue and counted as X(1).
Those of > 60 days age Y(1) are near-overdue and will appear in Month2 statement as X(2), if not banked by then.
Some of cheques arу pretty old and reside in the system for ages (over 1000 days), and therefore appear in each X monthly statement, while certain share of cheques in both X(n) and Y(n) will disappear from next month's X(n+1).
What would be the best logic for next month forecast basing on actual historical data? The most importand is X, but Y also welcome. It should be a forecast, as there is no next month's data as of yet.
The data is in SQL if it is relevant, but I need to understand the logic most, then I can generate code.

Answer 1

The logic below gets you the Number of Checks, Percent Overdue, Percent Near Overdue, and Percent All Other Checks counts for the current time period. Next step would be to create a query that does this for 3 or 12 months (or however many months you want to use to get projections). Then you trend out your total number of checks, and trend the percentages for each of the subcategories. You can use that to predict future amounts.
SELECT
Count(*) AS Number of Checks,
(SUM(CASE WHEN s.ageofcheck >= 90 THEN 1 ELSE 0 END))/Count(*) AS Number of Checks AS Percent Overdue Checks,
(SUM(CASE WHEN s.ageofcheck >= 60 AND s.ageofcheck < 90 THEN 1 ELSE 0 END))/Count(*) AS Number of Checks AS Percent Near Overdue Checks,
(SUM(CASE WHEN s.ageofcheck < 60 THEN 1 ELSE 0 END))/Count(*) AS Number of Checks AS Percent All Other Checks
FROM
(
SELECT
c.checknumber,
DATEDIFF(dd,GETDATE(),c.checkdate) AS ageofcheck
FROM
checks_table AS c
) AS s