Question

I have a table with a list of contract line items in the form of

CREATE TABLE contracts_lines (
  contract_id integer,
  product_id integer,
  contract_line_start datetime,
  contract_line_end datetime,
  amount float
)

What I would like to produce is a VIEW (or populate a table) that allows me to determine how much revenue I can expect each month – we have a simple rule that each line is recognized evenly over the term of the line (i.e. daily revenue = amount / (contract_line_end - contract_line_start) )

The VIEW will be used as the basis for a measure group in an SSAS cube, so its structure should be something like

montly_revenue_forecast (
  year int,
  month int,
  product_id int,
  contract_id int,
  amount float
)

I’d like to do 12 months worth of forecast, from the day it’s run and it will run through SSIS so I have access to its contstructs (like looping, etc.) I’d prefer not to have to write any stored procedures.

Any help is appreciated

Answer 1

What you need to do is first generate a sequence of dates for each contract/product and then use the sequence to group.

The most convenient way to generate a sequence is to use the rather poorly-documented spt_values table, like so:

WITH Contracts_CTE (contract_id, product_id, contract_day, amount) AS
(
    SELECT 
        cl.contract_id,
        cl.product_id,
        DATEADD(DAY, v.number, cl.contract_line_start),
        cl.amount / DATEDIFF(DAY, cl.contract_line_start, cl.contract_line_end)
    FROM contracts_lines cl
    CROSS JOIN master.dbo.spt_values v
    WHERE v.type = 'P'
    AND DATEADD(DAY, v.number, cl.contract_line_start) < cl.contract_line_end
)
SELECT
    DATEPART(YEAR, c.contract_day) AS contract_year,
    DATEPART(MONTH, c.contract_day) AS contract_month,
    c.product_id,
    c.contract_id,
    SUM(Amount) AS contract_amount
FROM Contracts_CTE c
GROUP BY
    c.product_id,
    c.contract_id,
    DATEPART(YEAR, c.contract_day),
    DATEPART(MONTH, c.contract_day)

Just a word of caution – the numbers in spt_values only go from 0 to 2047, so if you have contracts that are more than about 5 years long, then you’ll need to generate a longer sequence. Easiest way to do that is to CROSS JOIN the spt_values table to itself, i.e.:

SELECT (v1.number * 2048) + v2.number
FROM master.dbo.spt_values v1
CROSS JOIN master.dbo.spt_values v2
WHERE v1.type = 'P'
AND v2.type = 'P'
AND ((v1.number * 2048) + v2.number) < 100000

Obviously you’d have to integrate that into the above query, but if it’s really necessary then it shouldn’t be too hard.