Top 3 employee when emp changes its store - sql

Question

Table 1 = emp - (emp_id, store_id, start_dt, end_dt, amount)
Table 2 = Sales - (emp_id, product_id, sales_dateline, qty, amount)
Table 1 has data like this
emp_id store_id amount start_dt, end_dt
1 1 200 2/21/2019 10/21/2019
1 2 400 10/22/2019 12/31/2019
How can we find top 3 employees working on each store during Q4 of 2019 ( sales_dateline column).
Note - we need to consider amount from previous store also for each employee. store_id should be displayed in result set. please help.

Answer 1

This query gets the top 3 employees working in multiple stores during Q4 of 2019 based on the sum of the sales_dateline column of Table 2. The total sales per employee includes sales from each of the store_id's the employee worked in.
Step 1: (cte named 'q4_sales_per_employee_per_store'): Summarize the sales (sum_sales) per employee per store in q4 of 2019.
Step 2: (cte named 'q4_sales_per_multi_store_employee'): Summarize the aggregate sales (sum_sales) per
employee in q4 of 2019.
Step 3: (cte named 'q4_multi_store_employee_rankings'): Rank [row rank and dense rank (ties are given equal rank)] of aggregate sales (sum_sales) per employee in q4 of 2019.
Step 4: Select the top 3 (based on dense rank which could be more then 3 employees due to ties) multi-store employees with the highest sales (across all stores) JOIN'ed with per store sales.
Something like this
with
q4_sales_per_employee_per_store(emp_id, store_id, sum_qty, sum_amount) as (
select t1.emp_id, t1.store_id, sum(t2.qty), sum(t2.amount)
from Table1 t1
join Table2 t2 on t1.emp_id=t2.emp_id
where t2.sales_dateline>=cast('20191001' as date)
and t2.sales_dateline<cast('20200101' as date)
group by t1.emp_id, t1.store_id),
q4_sales_per_multi_store_employee(emp_id, sum_qty, sum_amount) as (
select qs.emp_id, sum(sum_qty), sum(sum_amount)
from q4_sales_per_employee_per_store
group by qs.emp_id
having count(*)>1)
q4_multi_store_employee_rankings(emp_id, sum_qty, sum_amount, sales_row, sales_rank) as (
select *,
row_number over (order by sum_amount desc) sales_row,
dense_rank over (order by sum_amount desc) sales_rank
from q4_sales_per_multi_store_employee)
select mser.sales_rank, mser.sales_row, mser.emp_id,
spe.store_id, spe.sum_qty, spe.sum_amount
from q4_multi_store_employee_rankings mser
join q4_sales_per_employee_per_store spe on mse.emp_id=spe.emp_id
where mser.sales_rank<=3
order by mser.sales_rank, mser.sales_row;

Answer 2

The I think your answer involves several logical steps. I'm calling them 'logical steps' because they will be combined together into one statement at the end (so the solution only has one query).
I'll start with adding variables for start and end dates rather than hardcoding them.
-- Period variables to define start and end of quarter
DECLARE #PeriodStart date = '20191001';
DECLARE #PeriodEnd date '20200101';
Note that when used, we use >= #PeriodStart but < #PeriodEnd so the enddate is midnight on the morning of the first day of the next quarter.
Then you need to work out the sales for each employee, regardless of store. I think you can do this simply with something like
SELECT emp_id, SUM(amount) AS TotalSales
FROM [sales] S
WHERE sales_dateline >= #PeriodStart AND sales_dateline < #PeriodEnd
GROUP BY emp_id;
Notes
I don't know how you want to do date filtering for sales - the name salesdateline field implies it's not simply a datetime, but possible a text field or reference to another table. While I've assumed a datetime in the above, I will leave any modifications up to you.
Assume 'amount' in the Sales table is a total for the sale for that product, not the unit price. If it's the unit price, you need to multiple this by the qty field e.g., SUM(amount * qty).
Then, you need to determine in which store each person was last working in during Q4 2019 (Oct 1 - Dec 31). This means if an employee started in Store 1 during the quarter, then movied to Store 2, their record will be for Store 2 (but take into account sales for Store 1).
The below gives an approach for this
It first finds a single record per Employee per Store if the employee has worked in that store within the relevant period (regardless of how long, or over how many different sessions).
It finds the record with the latest startdate within that period - that will be the store the Employee is assigned to for reporting
SELECT store_id, emp_id
FROM
(SELECT store_id, emp_id,
ROW_NUMBER() OVER (PARTITION BY store_id, emp_id ORDER BY start_dt) AS rn
FROM [emp] E
WHERE (start_dt >= #PeriodStart AND start_dt < #PeriodEnd) -- started in period
OR (end_dt >= #PeriodStart AND end_dt < #PeriodEnd) -- ended in period
OR (start_dt < #PeriodStart AND end_dt >= #PeriodEnd) -- worked through period
) AS AllEmpStores
WHERE AllEmpStores.rn = 1;
Note that the previous version of my answer, I just reported all stores each employee worked at within that period. Therefore if a great salesperson worked at 10 stores within the period, they could feasibly show in the top 3 list for every store. The code for that was
SELECT DISTINCT E.Store_ID, E.Emp_ID
FROM [Emp] E
WHERE (E.start_dt >= #PeriodStart AND E.start_dt < #PeriodEnd) -- started in period
OR (E.end_dt >= #PeriodStart AND E.end_dt < #PeriodEnd) -- ended in period
OR (E.start_dt < #PeriodStart AND E.end_dt >= #PeriodEnd); -- worked through period
Now you have the two datasets (sales per employee, and employees by store), you can then join the first to the second, so you have Store_ID, Emp_ID, and Total_Sales (across all stores).
At that point you just need the top 3 per store, which you can do with a windowed function.
DECLARE #PeriodStart date = '20191001';
DECLARE #PeriodEnd date '20200101';
WITH S_CTE AS
(SELECT emp_id, SUM(amount) AS TotalSales
FROM [sales] S
WHERE sales_dateline >= #PeriodStart AND sales_dateline < #PeriodEnd
GROUP BY emp_id;
),
E_CTE AS
(SELECT store_id, emp_id
FROM
(SELECT store_id, emp_id,
ROW_NUMBER() OVER (PARTITION BY store_id, emp_id ORDER BY start_dt) AS rn
FROM [emp] E
WHERE (start_dt >= #PeriodStart AND start_dt < #PeriodEnd)
OR (end_dt >= #PeriodStart AND end_dt < #PeriodEnd)
OR (start_dt < #PeriodStart AND end_dt >= #PeriodEnd)
) AS AllEmpStores
WHERE AllEmpStores.rn = 1
)
SELECT store_ID, emp_ID, TotalSales
FROM
(SELECT E_CTE.store_ID, E_CTE.emp_ID, S_CTE.TotalSales,
DENSE_RANK() OVER (PARTITION BY E_CTE.storeID ORDER BY S_CTE.TotalSales DESC) AS SalesRank
FROM E_CTE
INNER JOIN S_CTE ON E_CTE.emp_id = S_CTE.emp_id
) AS A
WHERE A.SalesRank <= 3
Edits/updates:
Now finds 'latest' store per employee only (rather than all stores) within period
Clarified filtering on sales
Added DENSE_RANK based on question comments
... And I just fixed a typo where E_CTE and S_CTE were the wrong way around.