Calculating YoY revenue when not all accounts existed last year - sql

Question

I am trying to calculate YoY revenue in Snowflake. I have 3 columns: account_id, activity_date, and arr. I've tried using a CTE to calculate, but the problem I have is that the account id's from today, might not all be the same from a year ago. I am trying to calculate this for each and every day and account. Here's what I have:
WITH CTE AS(SELECT account_id, account_name, activity_date, arr
FROM arr_base)
SELECT c1.*, c2.arr AS yoy_arr
FROM CTE c1
LEFT JOIN CTE c2 ON c1.account_id = c2.account_id AND c2.activity_date = dateadd(year, -1, c1.activity_date)
This approach ends up excluding some records, because not all accounts match. So the yoy_arr value is smaller than it should be. Any suggestions?
Edited to add sample data, current results and desired results:
Sample Data:
Account_ID
Activity_Date
ARR
A
Jan. 31, 2021
50
B
Jan. 31, 2021
40
A
Jan. 31, 2020
40
B
Jan. 31, 2020
35
C
Jan. 31, 2020
30
D
Jan. 31, 2020
30
Current Results:
Account_ID
Activity_Date
ARR
YOY_ARR
A
Jan. 31, 2021
50
40
B
Jan. 31, 2021
40
35
Desired Results:
Account_ID
Activity_Date
ARR
YOY_ARR
A
Jan. 31, 2021
50
40
B
Jan. 31, 2021
40
35
NULL
Jan. 31, 2021
NULL
60

Answer 1

This is an common Business Intelligence Challenge:
The solution to prevent having to do these calculations is usually to create a fact table.
In the fact table, one would persist every date for every account and be able to do quick yoy comparisons and there would be no missing values to content with.
you table would have records like
accountid, year, arr
C, 2019, 0
C, 2020, 30
C, 2021, 0
To give you a direct solution vs recommended strategy see the code here
create or replace table arr_base(account_id varchar, account_name varchar, activity_date date, arr number(32,4));
insert into arr_base (account_id, account_name, activity_date, arr)
values
('A','ACCOUNT A','2021-01-31',50)
,('B','ACCOUNT B','2021-01-31',40)
,('A','ACCOUNT A','2020-01-31',40)
,('B','ACCOUNT B','2020-01-31', 35)
,('C','ACCOUNT C','2020-01-31', 30)
,('D','ACCOUNT D','2020-01-31', 30)
;
set start_date = '2018-01-01'::timestamp_ltz;
set end_date = current_date();
set years = (select datediff(years,$start_date, $end_date) +1 );
with cte_accounts as (select distinct account_id, account_name from arr_base where activity_date between $start_date and $end_date )
,cte_dates as (select distinct year(activity_Date) as activity_year, activity_Date from arr_base where activity_date between $start_date and $end_date)
,cte_years as ( select seq4() + year($start_date) as txn_year from table(generator(rowcount=> $years)))
,cte_arr as (select account_id, account_name, year(activity_date) as activity_year, SUM( arr) as arr from arr_base where activity_date between $start_date and $end_date group by account_id, account_name, year(activity_date))
-- cartesian product
select
dim.activity_year
,dim.account_id
,dim.account_name
,f1.arr as arr_current_year
,f2.arr as arr_next_year
,case when f2.arr is not null and f1.arr is not null then ((f2.arr - f1.arr) / f1.arr) else null end as yoy_arr
from
(
select
a.account_id
,a.account_name
,d.activity_year
from
cte_accounts a cross join
cte_dates d
) as dim left outer join
cte_arr as f1 on dim.account_id = f1.account_id and dim.activity_year = f1.activity_year left outer join
cte_arr as f2 on dim.account_id = f2.account_id and (dim.activity_year+1) = f2.activity_year
order by dim.activity_year
,dim.account_id
;
See Results Below:
https://gist.github.com/umjohndacosta/784d5fa7a41a5e50066d925c92696c22

Answer 2

Alternative approach adding onto John's excellent answer:-)
select
*
from
(select
account_name
, year(ACTIVITY_DATE) year_arr
, to_char(ACTIVITY_DATE,'DD-MON') arr_date,arr
from arr_base )
pivot
( sum(arr) for year_arr in ('2021','2020' ))
create or replace table arr_base(account_id varchar, account_name varchar, activity_date date, arr number(32,4));
insert into arr_base (account_id, account_name, activity_date, arr)
values
('A','ACCOUNT A','2021-01-31',50)
,('B','ACCOUNT B','2021-01-31',40)
,('A','ACCOUNT A','2020-01-31',40)
,('B','ACCOUNT B','2020-01-31', 35)
,('C','ACCOUNT C','2020-01-31', 30)
,('D','ACCOUNT D','2020-01-31', 30)
;
select p.* from (select account_name, year(ACTIVITY_DATE) year_arr , to_char(ACTIVITY_DATE,'DD-MON') arr_date,arr from arr_base )
pivot(sum(arr) for year_arr in ('2021', '2020' )) as p