Current Day vs Previous Day in Table

Hi Everyone

Date	Sales	product
6/13/2016	21	A
6/12/2016	20	A

how to create a table like below by using date set above

Output

Product	Current_Day	previous_day	DoD
A	21	20	5%

Thanks in advance

Find more posts tagged with

Accepted answers

Gimli

Hello @iQuanti,

This can be accomplished through a SQL dataflow.
You will have two transforms.
Transform 1 will be to group your data down to a day view.

SELECT 

 `Date` 
,SUM(`Sales`) As 'Sales'  
,`Product`

FROM input_dataset

GROUP BY `Date`, `Product`

**Key notes

We can add more columns into the Select statement. We need to ensure they are added in the GROUP BY as well as the JOIN of the next transform
Your value column, (`sales` in this example) needs to have an aggregation. In this example I use SUM()
If your Date column is a Date/Time we will need to remove the time value. This can be accomplished by wrapping your column in DATE(). -- DATE(`Date`) -- This will also need to be done in the group by.

Transform 2 is were we join the data back onto itself.

SELECT 

 a.`Date`
,a.`Sales` AS 'Current_Day'
,b.`Sales` As 'Previous_Day'
,a.`Product`

FROM transform_data_1 a

LEFT JOIN transform_data_1 b
ON a.`Date` = DATE_ADD(b.`Date`, INTERVAL 1 DAY)
AND a.`Product` = b.`Product`

Your output of your dataflow would simply be:

SELECT * FROM transform_data_2

This also works for other date comparisons. If you wanted to compare today vs same day last week, we would only change the time key word:

ON a.`Date` = DATE_ADD(b.`Date`, INTERVAL 1 DAY)

To:

ON a.`Date` = DATE_ADD(b.`Date`, INTERVAL 1 WEEK)

Doing a total for a week compared to total for last week would require a change in our Grouping as well as in the JOIN conditions.

All comments

mdelorey

@iQuanti: Can you clarify what you're trying to do with the data? Right now it looks like the data has one row per day per product, and you're looking to get it to have one row per product that shows the current and previous day level as well as the percent change between current and previous day. Is that correct?

Gimli

Hello @iQuanti,

This can be accomplished through a SQL dataflow.
You will have two transforms.
Transform 1 will be to group your data down to a day view.

SELECT 

 `Date` 
,SUM(`Sales`) As 'Sales'  
,`Product`

FROM input_dataset

GROUP BY `Date`, `Product`

**Key notes

We can add more columns into the Select statement. We need to ensure they are added in the GROUP BY as well as the JOIN of the next transform
Your value column, (`sales` in this example) needs to have an aggregation. In this example I use SUM()
If your Date column is a Date/Time we will need to remove the time value. This can be accomplished by wrapping your column in DATE(). -- DATE(`Date`) -- This will also need to be done in the group by.

Transform 2 is were we join the data back onto itself.

SELECT 

 a.`Date`
,a.`Sales` AS 'Current_Day'
,b.`Sales` As 'Previous_Day'
,a.`Product`

FROM transform_data_1 a

LEFT JOIN transform_data_1 b
ON a.`Date` = DATE_ADD(b.`Date`, INTERVAL 1 DAY)
AND a.`Product` = b.`Product`

Your output of your dataflow would simply be:

SELECT * FROM transform_data_2

This also works for other date comparisons. If you wanted to compare today vs same day last week, we would only change the time key word:

ON a.`Date` = DATE_ADD(b.`Date`, INTERVAL 1 DAY)

To:

ON a.`Date` = DATE_ADD(b.`Date`, INTERVAL 1 WEEK)

Doing a total for a week compared to total for last week would require a change in our Grouping as well as in the JOIN conditions.

iQuanti

I really appreciate your help in resolving the issue ilikenno. Could you please solve the one more request from me, how to add last 30 days average in the same table by using SQL Dataflow.

Date	Current_Day	previous_day	last 30 days average
6/1/2015	20	22	25

Regards,

iQuanti

kshah008

@iQuanti, feel free to open up a new thread for your new question for better exposure ?

canioacaputo

I really like the way you did this, next is capturing the difference from day to day so all I did was say

(a.users - b.users) as 'Change' and I get my day to day change

jkimmey23

Hi,

When using this solution I am running into the issue with the "1" in the "interval 1 day" part of the formula and seeing the following error message - "The database reported a syntax error. ERROR: syntax error at or near "'Date'" Position: 83". Is there a special way to write out this part to avoid this error?