{ "id": 8287, "title": "Offer Quantity and Value", "url": "/competitions/acquire-valued-shoppers-challenge/discussion/8287", "author_name": "", "post_date": "2014-05-26T22:18:34.523Z", "votes": 3, "comment_count": 16, "views": 5151, "content": "

Hi all,

\n

Examining the data and trying to exploit all transactions (not only if customer X has bought the same product/brand/company) I find the explanation of some of the data somewhat confusing. 

\n

Mainly, what is the meaning of Offer value and quantity? According to the specifications in offers.csv:

\n

quantity - The number of units one must purchase to get the discount

\n

I really find this confusing, because in transactions.csv there are transactions of the same product with different 'productsize' and 'productquantity'. So, I'm not sure how to interpret this quantity.  For example, I found that product (3203, 13474, 106414464 / category, brand, company) has 131492 transactions registered with the following product sizes:( 24, 20, 10, 8, 6, 5). You can assume the sizes are in Liters, but it doesn't really matter

\n

offervalue - The dollar value of the offer

\n

This field is related to the other, as I assume this is the total amount of money a customer will have as a discount for buying a particular product. But, how does that relate with the quantity/size of the transactions? 

\n

Following the same example as before, the offer for the product mentioned says that there its value of 0.75 buying a minimum of 1 unit. There are many problems here:

\n

- Is it a 0.75$ dollars discount regardless of the size bought (5L or 8L, for example)? That doesn't make much sense to me

\n

- Is  it 0.75$ dollars for each unit (Liters) purchased? In that case, examining prices, some of them would be negative, which doesn't make much sense either

\n

Is someone in the same situation? 

\n\n

Thanks

", "messages": [ { "id": "47102", "postDate": "05/26/2014 22:18:34", "content": "

Hi all,

\n

Examining the data and trying to exploit all transactions (not only if customer X has bought the same product/brand/company) I find the explanation of some of the data somewhat confusing. 

\n

Mainly, what is the meaning of Offer value and quantity? According to the specifications in offers.csv:

\n

quantity - The number of units one must purchase to get the discount

\n

I really find this confusing, because in transactions.csv there are transactions of the same product with different 'productsize' and 'productquantity'. So, I'm not sure how to interpret this quantity.  For example, I found that product (3203, 13474, 106414464 / category, brand, company) has 131492 transactions registered with the following product sizes:( 24, 20, 10, 8, 6, 5). You can assume the sizes are in Liters, but it doesn't really matter

\n

offervalue - The dollar value of the offer

\n

This field is related to the other, as I assume this is the total amount of money a customer will have as a discount for buying a particular product. But, how does that relate with the quantity/size of the transactions? 

\n

Following the same example as before, the offer for the product mentioned says that there its value of 0.75 buying a minimum of 1 unit. There are many problems here:

\n

- Is it a 0.75$ dollars discount regardless of the size bought (5L or 8L, for example)? That doesn't make much sense to me

\n

- Is  it 0.75$ dollars for each unit (Liters) purchased? In that case, examining prices, some of them would be negative, which doesn't make much sense either

\n

Is someone in the same situation? 

\n\n

Thanks

", "rawMarkdown": "", "votes": null }, { "id": "47306", "postDate": "05/30/2014 05:26:58", "content": "

For me seems that product size is thrash field, 'cause we can see in transactions many products with widely varying product size for SAME price. That seems to me as it can't be used..

", "rawMarkdown": "", "votes": null }, { "id": "47417", "postDate": "05/31/2014 21:46:31", "content": "

That's what I observed too.

\n

I tried to compare prices for the same category products with different brands and company, but it seems rather difficult. I thought it would be an important factor when choosing one product over the other. Apparently we can't easily exploit this information

", "rawMarkdown": "", "votes": null }, { "id": "48422", "postDate": "06/01/2014 05:17:01", "content": "

What was problem for you? (Our team made this task, so I can help you with figuring out easiest way)

", "rawMarkdown": "", "votes": null }, { "id": "48443", "postDate": "06/01/2014 20:21:40", "content": "

For example, for the product (9115, 93904, 108500080) (category, brand, company)

\n

I got these prices for 0.2 and 0.05 quantities of product:

\n

0.2, [ 4. , 4.5 , 5.99 ]
0.05, [ 0.89, 0.99, 0.99, 1. , 1.49]

\n\n

There are more quantities, there are more prices also. But this is just an example for one particular product. 

", "rawMarkdown": "", "votes": null }, { "id": "48445", "postDate": "06/01/2014 20:50:51", "content": "

That data makes sense.

\n

Look at real-world examples:

\n

http://www.coupons.com/

\n

"Save $1 on any purchase of (two) Ziploc brand bags."

\n

It doesn't matter the size of the container purchased. In this case, you will save $1 if you buy one box of 100 count and one box of 500 count, or just 2 boxes of 100 count.

", "rawMarkdown": "", "votes": null }, { "id": "48446", "postDate": "06/01/2014 21:14:21", "content": "

That's what I'm doing right now. Although having the same discount for two completely different sizes seems odd to me.

", "rawMarkdown": "", "votes": null }, { "id": "48449", "postDate": "06/01/2014 21:33:18", "content": "

I think it's ok just to know price per unit. So you can divide amount of money of quntity of item, and then find mean price of median, what you prefer.

", "rawMarkdown": "", "votes": null }, { "id": "48519", "postDate": "06/02/2014 21:11:32", "content": "

Yes,

\n

I was using the median, it seemed a reasonable approximation. 

\n

First experiments show that many offered products are way expensive than the others. Did you find anything similar?

", "rawMarkdown": "", "votes": null }, { "id": "48548", "postDate": "06/03/2014 07:01:43", "content": "

I didn't watch concrete offer prices, just try to use them. I will rethink this part on this week so I'll write you about what I find

", "rawMarkdown": "", "votes": null }, { "id": "48911", "postDate": "06/09/2014 16:28:09", "content": "

Hi everybodi,

\n\n

is logical that there are different offers exactly the same? For example;

\n\n

offer        categ    q.   company   offervalue brand

\n

1198271  5558    1   107120272    1.5          5072
1198272  5558    1   107120272    1.5          5072
1198273  5558    1   107120272    1.5          5072
1198274  5558    1   107120272    1.5          5072
1198275  5558    1   107120272    1.5          5072

\n\n

regards!

", "rawMarkdown": "", "votes": null }, { "id": "48913", "postDate": "06/09/2014 16:43:28", "content": "

Hi,

\n

I won't be too worried about that, real company data usually is messy, maybe they don't have time to check whether the offer already exists so they create a new one exactly the same, it's very possible.

\n\n

Mike

", "rawMarkdown": "", "votes": null }, { "id": "48918", "postDate": "06/09/2014 17:37:11", "content": "

You found information that you think is useful, so use it -- make better predictor ;)

", "rawMarkdown": "", "votes": null }, { "id": "48919", "postDate": "06/09/2014 17:37:54", "content": "

Could be the offers were in different markets, different chains, or launched at different times. I'd look into that if you think it's important.

", "rawMarkdown": "", "votes": null }, { "id": "48925", "postDate": "06/09/2014 21:39:40", "content": "

In the offers table, it's basically saying the customer can get e.g. $1 discount when purchasing an item in category X, brand Y. This is a one-time event.

\n

It's not uncommon for the offer to be valid for more than one product size, e.g. the 8 oz, 12 oz or 16 oz package size. The company making the offer simply wants you to try the product and doesn't care which size you buy (but you can pick only one).

\n

As for the prices in the transaction file, these are actual prices paid. It's common to have differing prices by market or time. Some promotions are from the manufacturer, and will exist in several chains at once, but sometimes a chain (or even an individual store) may make a promotional offer, and so prices can vary for the same item at different chains.

", "rawMarkdown": "", "votes": null }, { "id": "49016", "postDate": "06/11/2014 17:34:15", "content": "

can we use the median as a new feature? Is this a good idea?  Thank you all.

", "rawMarkdown": "", "votes": null }, { "id": "49019", "postDate": "06/11/2014 18:09:32", "content": "

In other thread there was a person who said that median values in addtition to means helped him a lot

", "rawMarkdown": "", "votes": null } ], "comments": [ { "id": 47306, "author_name": "dremovd", "author_url": "", "post_date": "05/30/2014 05:26:58", "content": "

For me seems that product size is thrash field, 'cause we can see in transactions many products with widely varying product size for SAME price. That seems to me as it can't be used..

", "votes": null, "replies": [] }, { "id": 47417, "author_name": "guillempalou", "author_url": "", "post_date": "05/31/2014 21:46:31", "content": "

That's what I observed too.

\n

I tried to compare prices for the same category products with different brands and company, but it seems rather difficult. I thought it would be an important factor when choosing one product over the other. Apparently we can't easily exploit this information

", "votes": null, "replies": [] }, { "id": 48422, "author_name": "dremovd", "author_url": "", "post_date": "06/01/2014 05:17:01", "content": "

What was problem for you? (Our team made this task, so I can help you with figuring out easiest way)

", "votes": null, "replies": [] }, { "id": 48443, "author_name": "guillempalou", "author_url": "", "post_date": "06/01/2014 20:21:40", "content": "

For example, for the product (9115, 93904, 108500080) (category, brand, company)

\n

I got these prices for 0.2 and 0.05 quantities of product:

\n

0.2, [ 4. , 4.5 , 5.99 ]
0.05, [ 0.89, 0.99, 0.99, 1. , 1.49]

\n\n

There are more quantities, there are more prices also. But this is just an example for one particular product. 

", "votes": null, "replies": [] }, { "id": 48445, "author_name": "inversion", "author_url": "", "post_date": "06/01/2014 20:50:51", "content": "

That data makes sense.

\n

Look at real-world examples:

\n

http://www.coupons.com/

\n

"Save $1 on any purchase of (two) Ziploc brand bags."

\n

It doesn't matter the size of the container purchased. In this case, you will save $1 if you buy one box of 100 count and one box of 500 count, or just 2 boxes of 100 count.

", "votes": null, "replies": [] }, { "id": 48446, "author_name": "guillempalou", "author_url": "", "post_date": "06/01/2014 21:14:21", "content": "

That's what I'm doing right now. Although having the same discount for two completely different sizes seems odd to me.

", "votes": null, "replies": [] }, { "id": 48449, "author_name": "dremovd", "author_url": "", "post_date": "06/01/2014 21:33:18", "content": "

I think it's ok just to know price per unit. So you can divide amount of money of quntity of item, and then find mean price of median, what you prefer.

", "votes": null, "replies": [] }, { "id": 48519, "author_name": "guillempalou", "author_url": "", "post_date": "06/02/2014 21:11:32", "content": "

Yes,

\n

I was using the median, it seemed a reasonable approximation. 

\n

First experiments show that many offered products are way expensive than the others. Did you find anything similar?

", "votes": null, "replies": [] }, { "id": 48548, "author_name": "dremovd", "author_url": "", "post_date": "06/03/2014 07:01:43", "content": "

I didn't watch concrete offer prices, just try to use them. I will rethink this part on this week so I'll write you about what I find

", "votes": null, "replies": [] }, { "id": 48911, "author_name": "mike64", "author_url": "", "post_date": "06/09/2014 16:28:09", "content": "

Hi everybodi,

\n\n

is logical that there are different offers exactly the same? For example;

\n\n

offer        categ    q.   company   offervalue brand

\n

1198271  5558    1   107120272    1.5          5072
1198272  5558    1   107120272    1.5          5072
1198273  5558    1   107120272    1.5          5072
1198274  5558    1   107120272    1.5          5072
1198275  5558    1   107120272    1.5          5072

\n\n

regards!

", "votes": null, "replies": [] }, { "id": 48913, "author_name": "mikedemurtas", "author_url": "", "post_date": "06/09/2014 16:43:28", "content": "

Hi,

\n

I won't be too worried about that, real company data usually is messy, maybe they don't have time to check whether the offer already exists so they create a new one exactly the same, it's very possible.

\n\n

Mike

", "votes": null, "replies": [] }, { "id": 48918, "author_name": "dremovd", "author_url": "", "post_date": "06/09/2014 17:37:11", "content": "

You found information that you think is useful, so use it -- make better predictor ;)

", "votes": null, "replies": [] }, { "id": 48919, "author_name": "inversion", "author_url": "", "post_date": "06/09/2014 17:37:54", "content": "

Could be the offers were in different markets, different chains, or launched at different times. I'd look into that if you think it's important.

", "votes": null, "replies": [] }, { "id": 48925, "author_name": "rsteger", "author_url": "", "post_date": "06/09/2014 21:39:40", "content": "

In the offers table, it's basically saying the customer can get e.g. $1 discount when purchasing an item in category X, brand Y. This is a one-time event.

\n

It's not uncommon for the offer to be valid for more than one product size, e.g. the 8 oz, 12 oz or 16 oz package size. The company making the offer simply wants you to try the product and doesn't care which size you buy (but you can pick only one).

\n

As for the prices in the transaction file, these are actual prices paid. It's common to have differing prices by market or time. Some promotions are from the manufacturer, and will exist in several chains at once, but sometimes a chain (or even an individual store) may make a promotional offer, and so prices can vary for the same item at different chains.

", "votes": null, "replies": [] }, { "id": 49016, "author_name": "potato", "author_url": "", "post_date": "06/11/2014 17:34:15", "content": "

can we use the median as a new feature? Is this a good idea?  Thank you all.

", "votes": null, "replies": [] }, { "id": 49019, "author_name": "dremovd", "author_url": "", "post_date": "06/11/2014 18:09:32", "content": "

In other thread there was a person who said that median values in addtition to means helped him a lot

", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "47102": "", "47306": "", "47417": "", "48422": "", "48443": "", "48445": "", "48446": "", "48449": "", "48519": "", "48548": "", "48911": "", "48913": "", "48918": "", "48919": "", "48925": "", "49016": "", "49019": "" }, "source": "meta" }