BigQuery 中的逐年逐月比较和本月至今

Question

Edit: @shawnt00 has the correct answer. Thank you very much!

I am having trouble accurately doing a year over year comparison by month but at any point during the month. For example for August 2022 vs 2021, I want to compare August 1 - August 25, rather than full month of August 2021.

I am also using a daily date field.

I want the final result to basically be:

Product_ID, Year, Month, PY_Sales, CY_Sales

Edit: I have daily totals. Some products do have not sales on certain days though:

product_id	sale_date	units
1	2021-01-01	5
2	2021-01-02	4
...	...	...
1	2021-06-05	2
2	2022-01-06	1
2	2022-08-15	9

This is the code I have, but it doesn't do MTD. So 2021 August is the entire month of August and I want it the same dates for 2022. I used this code because some products do not have sales on certain months.

    WITH cte AS 
(
SELECT 
     PRODUCT_ID,
     EXTRACT(YEAR FROM SALE_DATE) AS Year,
     EXTRACT(MONTH FROM SALE_DATE) AS Month,
     CONCAT(EXTRACT(YEAR FROM SALE_DATE), '-',EXTRACT(MONTH FROM SALE_DATE)) AS Year_Month,
     SUM(Units) AS Units
FROM data
WHERE Product_ID = 1
AND DATE(SALE_DATE) >= '2019-01-01'
GROUP BY 1, 2, 3
),

    diff AS 
( 
SELECT 
     COALESCE(c.PRODUCT_ID, p.PRODUCT_ID) AS Product_ID,
     COALESCE(c.Year, p.Year + 1) AS Year,
     COALESCE(c.Month, p.Month) AS Month,
     IFNULL(c.Units, 0) AS Current_Units,
     IFNULL(p.Units, 0) AS Previous_Units,
     NULLIF(((IFNULL(c.Units, 0) - IFNULL(p.Units,0)) / p.Units),0) * 100 AS Percent_Change
FROM CTE c
FULL OUTER JOIN CTE p ON c.PRODUCT_ID = p.PRODUCT_ID AND c.Year = p.Year + 1 AND c.Month = p.Month
WHERE c.Year <= EXTRACT(YEAR FROM CURRENT_DATE())
ORDER BY 2, c.Year, c.Month
)
SELECT *
FROM diff
--This is to avoid dividing by 0
WHERE diff.Previous_Units > 0
--AND Percent_Change <= -.5

Answer 1

You could just roll up two different monthly totals and then switch for the current month comparison:

with agg as (
    select
        PRODUCT_ID,
        extract(year from SALE_DATE) as yr,
        extract(month from SALE_DATE) as mth,
        sum(Units) as Units,
        sum(case when extract(day from SALE_DATE) <= extract(day from current_date())
                 then Units end) as UnitsMTD
    from data
    where date(SALE_DATE) >= '2019-01-01' -- one year before report output
    group by 1, 2, 3
)
select c.Yr, c.Mth, c.PRODUCT_ID,
    case when Yr  = extract(year  from current_date())
          and Mth = extract(month from current_date())
         then (c.UnitsMTD - p.UnitsMTD) / p.UnitsMTD
         else (c.Units    - p.Units   ) / p.Units
    end as Percent_Change
from agg c left outer join agg p
    on p.Product_ID = c.Product_ID and p.Yr = c.Yr - 1 and p.Mth = c.Mth
order by c.Yr, c.Mth, c.PRODUCT_ID;

Note my earlier comment about leap years. This will treat February 28 of the year following a leap year as an "MTD" month. You might need to handle that differently inside the case expression.