• analyzing supermarket's business performance after new packaging program launched
• using MS. SQL SERVER STUDIO

SQL project/case study from : https://8weeksqlchallenge.com/case-study-5/

steps:

1. import data_mart_weekly_sales.csv into SQL SERVER

2. data cleaning -> create new table: clean_weekly_sales
   

   clean_weekly_sales

   select region, platform, segment, customer_type, transactions, sales,
   SUBSTRING(week_date,2, CHARINDEX('/', week_date, 1)-2)  as day,
   SUBSTRING(week_date, CHARINDEX('/', week_date, 1)+1, 1) as month_number,
   concat('20', SUBSTRING(week_date, CHARINDEX('/', week_date, CHARINDEX('/', week_date)+1)+1, 2)) as calendar_year,
   DATEFROMPARTS(concat('20', SUBSTRING(week_date, CHARINDEX('/', week_date, CHARINDEX('/', week_date)+1)+1, 2)),
   SUBSTRING(week_date, CHARINDEX('/', week_date, 1)+1, 1),
   SUBSTRING(week_date,2, CHARINDEX('/', week_date, 1)-2)) as date_full,
   case
   when SUBSTRING(week_date,2, CHARINDEX('/', week_date, 1)-2) between 1 and 7 then 1
   when SUBSTRING(week_date,2, CHARINDEX('/', week_date, 1)-2) between 8 and 14 then 2
   when SUBSTRING(week_date,2, CHARINDEX('/', week_date, 1)-2) between 15 and 21 then 3
   else 4
   end as week_number,
   case
   when SUBSTRING(segment, 2, 1) = 'C' then 'Couples'
   when SUBSTRING(segment, 2, 1) = 'F' then 'Families'
   else 'unknown'
   end as demographic,
   case
   when SUBSTRING(segment, 3, 1) = '1' then 'Young Adults'
   when SUBSTRING(segment, 3, 1) = '2' then 'Middle Aged'
   when SUBSTRING(segment, 3, 1) = '3' or SUBSTRING(segment, 3, 1) = '4' then 'Retirees'
   else 'unknown'
   end as age_band,
   round(sales*100.0/transactions, 2) as avg_transaction
   into clean_weekly_sales
   from data_mart_weekly_sales;
   

   

3. data exploration:

   1. What day of the week is used for each week_date value?

      day_used

       select distinct date_full, DAY(date_full) as day_used, DATENAME(weekday, date_full) as day_name
       from clean_weekly_sales
       order by date_full;

      Monday

      

   2. What range of week numbers are missing from the dataset?

      week numbers

       select distinct week_number
       from clean_weekly_sales
       order by week_number;

      none is missing

      

   3. How many total transactions were there for each year in the dataset?

      total transactions for each year

       select calendar_year, SUM(transactions) as total_trx
       from clean_weekly_sales
       group by calendar_year;
       

      

   4. What is the total sales for each region for each month?

      total sales for each region, each month

       with satu as (select *
       from (select region, month_number, CAST(sales as bigint) as sales
           from clean_weekly_sales) s
       pivot(
           sum(sales)
           for region in (['AFRICA'], ['ASIA'], ['CANADA'], ['EUROPE'], ['OCEANIA'], ['SOUTH AMERICA'], ['USA'])
       ) pivot_table)
      select *
      from satu
      order by month_number;
      

      

   5. What is the total count of transactions for each platform?

      total count of transactions for each platform

       select platform, COUNT(transactions) as total_cnt_trx
       from clean_weekly_sales
       group by platform;
       

      

   6. What is the percentage of sales for Retail vs Shopify for each month?

      percentage of sales for Retail vs Shopify for each month

       with satu as (select calendar_year, month_number, platform, sum(CAST(sales as bigint)) as total_sales
       from clean_weekly_sales
       group by calendar_year, month_number, platform
       ), 
       
      
       dua as (select *, 
       SUM(total_sales) over(partition by calendar_year) as total_sales_all, 
       round(total_sales * 100.0 / SUM(total_sales) over(partition by calendar_year, month_number), 2) as pct_sales_per_month
       from satu)
       
      
       select *
       from (select calendar_year, platform, month_number, pct_sales_per_month from dua) s
       pivot (
           max(pct_sales_per_month)
           for calendar_year in ([2018], [2019], [2020])
       ) pvt        
       

      

   7. What is the percentage of sales by demographic for each year in the dataset?

      percentage of sales by demographic for each year

       with satu as (select calendar_year, demographic, sum(CAST(sales as bigint)) as total_sales
       from clean_weekly_sales
       group by calendar_year, demographic),
       
      
       dua as (select *, SUM(total_sales) over(partition by calendar_year) as total_sales_all,
       round(total_sales * 100.0 / SUM(total_sales) over(partition by calendar_year), 2) as pct_demographic_per_year
       from satu)
       
              
       select *
       from (select demographic, calendar_year, pct_demographic_per_year from dua) s
       pivot(
           max(pct_demographic_per_year)
           for calendar_year in ([2018], [2019], [2020])
       ) pvt
       

      

   8. Which age_band and demographic values contribute the most to Retail sales?

      age_band and demographic values contribute the most to Retail sales

      retirees-families and retirees-couples

       select platform, age_band, demographic, sum(CAST(sales as bigint)) as total_sales, RANK() over(order by sum(CAST(sales as bigint)) desc) as ranking
       from clean_weekly_sales
       where platform = '''Retail''' and age_band <> 'unknown'
       group by platform, age_band, demographic;
       

      

   9. Can we use the avg_transaction column to find the average transaction size for each year for Retail vs Shopify? If not - how would you calculate it instead?

      average transaction size for each year for Retail vs Shopify

       select calendar_year, platform, avg(CAST(sales as bigint)) as avg_trx
       from clean_weekly_sales
       group by calendar_year, platform
       order by calendar_year, platform;        
       

      

4. analyzing data:

   1. What is the total sales for the 4 weeks before and after 2020-06-15? What is the growth or reduction rate in actual values and percentage of sales?

      total sales for the 4 weeks before and after 2020-06-15

       with satu as (select *,
       case
       when date_full >= '2020-06-15' then 'after' else 'before'
       end as new_packaging_date
       from clean_weekly_sales),
      
      total_sales_4_weeks_before as(
      select sum(CAST(sales as bigint)) as total_sales_before
      from satu
      where date_full between DATEADD(week, -4, '2020-06-15') and '2020-06-15'),
      
      total_sales_4_weeks_after as(
      select sum(CAST(sales as bigint)) as total_sales_after
      from satu
      where date_full between '2020-06-15' and DATEADD(week, 4, '2020-06-15'))
      
      select *,
      case
      when total_sales_4_weeks_before.total_sales_before > total_sales_4_weeks_after.total_sales_after then 'before is more trx' else 'after is more trx'
      end as status,
      total_sales_4_weeks_before.total_sales_before - total_sales_4_weeks_after.total_sales_after as diff
      from total_sales_4_weeks_before, total_sales_4_weeks_after;
      
      

      
      

      more trx before new program launched, in timeline 4 weeks before and 4 weeks after

      the difference: 10.973.134

      

   2. What about the entire 12 weeks before and after?

      total sales for the 12 weeks before and after 2020-06-15

       with satu as (select *,
       case
       when date_full >= '2020-06-15' then 'after' else 'before'
       end as new_packaging_date
       from clean_weekly_sales),
      total_sales_12_weeks_before as(
      select sum(CAST(sales as bigint)) as total_sales_before
      from satu
      where date_full between DATEADD(week, -12, '2020-06-15') and '2020-06-15'),
      total_sales_12_weeks_after as(
      select sum(CAST(sales as bigint)) as total_sales_after
      from satu
      where date_full between '2020-06-15' and DATEADD(week, 12, '2020-06-15'))
      select *,
      case
      when total_sales_12_weeks_before.total_sales_before > total_sales_12_weeks_after.total_sales_after then 'before is more trx' else 'after is more trx'
      end as status,
      total_sales_12_weeks_before.total_sales_before - total_sales_12_weeks_after.total_sales_after as diff
      from total_sales_12_weeks_before, total_sales_12_weeks_after
      
      

      
      

      more trx before new program launched, in timeline 12 weeks before and 12 weeks after

      the difference got more bigger for 'before' status: 722.350.742

      

   3. How do the sale metrics for these 2 periods before and after compare with the previous years in 2018 and 2019?

      compare with the previous years in 2018 and 2019

       select calendar_year, sum(CAST(sales as bigint)) as total_sales
       from clean_weekly_sales
       where calendar_year in (2018, 2019)
       group by calendar_year;     
       

      
      

      with total_sales before new program:

           - 2018 around 13 million
           - 2019 around 14 million
           - 4 weeks before and
           - 12 weeks before
      

      all show bigger total_sales compared to total_sales after new program launched

      

5. more question: Which areas of the business have the highest negative impact in sales metrics performance in 2020 for the 12 week before and after period?

   1. region

      
      

      region with highest negative impact

      with satu as (select *,
      case
      when date_full >= '2020-06-15' then 'after' else 'before'
      end as new_packaging_date
      from clean_weekly_sales),
      
      total_sales_12_weeks_before as(
      select region, sum(CAST(sales as bigint)) as total_sales_before
      from satu
      where date_full between DATEADD(week, -12, '2020-06-15') and '2020-06-15'
      group by region),
      
      total_sales_12_weeks_after as(
      select region, sum(CAST(sales as bigint)) as total_sales_after
      from satu
      where date_full between '2020-06-15' and DATEADD(week, 12, '2020-06-15')
      group by region)
      
      select b.region,
      b. total_sales_before,
      a.total_sales_after,
      (b.total_sales_before - a.total_sales_after) as diff,
      round((b.total_sales_before - a.total_sales_after) *100.0 / b. total_sales_before, 2) as pct_impact
      from total_sales_12_weeks_before b
      join total_sales_12_weeks_after a
      on b.region = a.region
      order by (b.total_sales_before - a.total_sales_after) *100.0 / b. total_sales_before desc;
      
      

      
      

      the region with highest negative impact is asia

      

   2. platform

      
      

      platform with highest negative impact

      with satu as (select *,
      case
      when date_full >= '2020-06-15' then 'after' else 'before'
      end as new_packaging_date
      from clean_weekly_sales),
      
      total_sales_12_weeks_before as(
      select platform, sum(CAST(sales as bigint)) as total_sales_before
      from satu
      where date_full between DATEADD(week, -12, '2020-06-15') and '2020-06-15'
      group by platform),
      
      total_sales_12_weeks_after as(
      select platform, sum(CAST(sales as bigint)) as total_sales_after
      from satu
      where date_full between '2020-06-15' and DATEADD(week, 12, '2020-06-15')
      group by platform)
      
      select b.platform,
      b. total_sales_before,
      a.total_sales_after,
      (b.total_sales_before - a.total_sales_after) as diff,
      round((b.total_sales_before - a.total_sales_after) *100.0/ b. total_sales_before, 2) as pct_impact
      from total_sales_12_weeks_before b
      join total_sales_12_weeks_after a
      on b.platform = a.platform
      order by round((b.total_sales_before - a.total_sales_after) *100.0/ b. total_sales_before, 2) desc;
      
      

      
      

      the platform with highest negative impact is retail

      

   3. age_band

      
      

      age_band with highest negative impact

      with satu as (select *,
      case
      when date_full >= '2020-06-15' then 'after' else 'before'
      end as new_packaging_date
      from clean_weekly_sales),
      
      total_sales_12_weeks_before as(
      select age_band, sum(CAST(sales as bigint)) as total_sales_before
      from satu
      where date_full between DATEADD(week, -12, '2020-06-15') and '2020-06-15'
      group by age_band),
      
      total_sales_12_weeks_after as(
      select age_band, sum(CAST(sales as bigint)) as total_sales_after
      from satu
      where date_full between '2020-06-15' and DATEADD(week, 12, '2020-06-15')
      group by age_band)
      
      select b.age_band,
      b. total_sales_before,
      a.total_sales_after,
      (b.total_sales_before - a.total_sales_after) as diff,
      round((b.total_sales_before - a.total_sales_after) *100.0/ b. total_sales_before, 2) as pct_impact
      from total_sales_12_weeks_before b
      join total_sales_12_weeks_after a
      on b.age_band = a.age_band
      where b.age_band <> 'unknown'
      order by round((b.total_sales_before - a.total_sales_after) *100.0/ b. total_sales_before, 2) desc;
      
      

      
      

      the age_band with highest negative impact is middle aged

      

   4. demographic

      
      

      demographic with highest negative impact

      with satu as (select *,
      case
      when date_full >= '2020-06-15' then 'after' else 'before'
      end as new_packaging_date
      from clean_weekly_sales),
      
      total_sales_12_weeks_before as(
      select demographic, sum(CAST(sales as bigint)) as total_sales_before
      from satu
      where date_full between DATEADD(week, -12, '2020-06-15') and '2020-06-15'
      group by demographic),
      
      total_sales_12_weeks_after as(
      select demographic, sum(CAST(sales as bigint)) as total_sales_after
      from satu
      where date_full between '2020-06-15' and DATEADD(week, 12, '2020-06-15')
      group by demographic)
      
      select b.demographic,
      b. total_sales_before,
      a.total_sales_after,
      (b.total_sales_before - a.total_sales_after) as diff,
      round((b.total_sales_before - a.total_sales_after) *100.0/ b. total_sales_before, 2) as pct_impact
      from total_sales_12_weeks_before b
      join total_sales_12_weeks_after a
      on b.demographic = a.demographic
      where b.demographic <> 'unknown'
      order by round((b.total_sales_before - a.total_sales_after) *100.0/ b. total_sales_before, 2) desc;
      
      

      
      

      the demographic with highest negative impact is families

      

   5. customer_type

      
      

      customer type with highest negative impact

      with satu as (select *,
      case
      when date_full >= '2020-06-15' then 'after' else 'before'
      end as new_packaging_date
      from clean_weekly_sales),
      
      total_sales_12_weeks_before as(
      select customer_type, sum(CAST(sales as bigint)) as total_sales_before
      from satu
      where date_full between DATEADD(week, -12, '2020-06-15') and '2020-06-15'
      group by customer_type),
      
      total_sales_12_weeks_after as(
      select customer_type, sum(CAST(sales as bigint)) as total_sales_after
      from satu
      where date_full between '2020-06-15' and DATEADD(week, 12, '2020-06-15')
      group by customer_type)
      
      select b.customer_type,
      b. total_sales_before,
      a.total_sales_after,
      (b.total_sales_before - a.total_sales_after) as diff,
      round((b.total_sales_before - a.total_sales_after) *100.0/ b. total_sales_before, 2) as pct_impact
      from total_sales_12_weeks_before b
      join total_sales_12_weeks_after a
      on b.customer_type = a.customer_type
      order by round((b.total_sales_before - a.total_sales_after) *100.0/ b. total_sales_before, 2) desc;
      
      

      
      

      the customer type with highest negative impact is guest