Sales & Promotions Data Analysis

1. Data Analyzing & Understanding

All the campaigns were added to the sales data in order to make sales analyzes for the campaign periods.

First of all, an overview of the number of campaigns and sales amount was taken on a monthly basis (returns were excluded to see only sales, returns will be examined separately).

Sales

Return

The first striking details here;

• In the 3rd month, campaign usage and sales amount increased in parallel

• Even though there was no campaign in the 4th month, there were more sales than the 2 months with the campaign (February and June).

• While the number of returns increases with the campaigns in the 2nd and 3rd months, the number of returns is not parallel to the number of sales and decreases in the 5th month. This seems important in this scenario. While examining the decrease in return in the 5th month, the campaign held in this month can be evaluated separately in this sense.

Sales quantities and campaign usage & campaign return by days of the week

Sales

Return

• When we look at sales and campaign sales according to the days of the week, we see an average trend on Monday, Tuesday and Wednesday, an increasing trend on Thursday, Friday and Sunday, and a decreasing trend on Saturday.

• Although the day with the highest campaign usage is Sunday, the number of sales is lower than Friday. Here, we can perhaps comment that the goal of people shopping on Sunday is to benefit from the campaign, and that they are generally not inclined to shop outside of the campaign.

• In order to observe whether the campaigns had an effect on the sales distribution according to the days of the week, the sales of the periods without campaigns were also added to the chart. From this graph, we observe that the campaigns have no effect on the daily trend.

• The Friday, Saturday and Sunday return trend looks similar to the selling trend. Although the number of return transactions is higher on Sunday during campaign periods, the number of returns is almost similar on other days except Saturday. As a comment, orders purchased on other days may also be returned on Sunday.

• When we look at the campaign period returns, we can see that the daily trend is the same and does not affect the trend.

Sales quantities and campaign usage & campaign return by week of the year

Sales

Return

There were sudden increases in sales during the campaign periods compared to the previous week. The return increase for Promo1 continues the next week after the campaign. The most sudden increase in sales was in the Promo2 period.

Sales distribution week of year

2. Segmentation

Item segmentation

Pareto method was used for product classification.

In order to classify the products as Fast Item, Medium Item and Slow Item, firstly non-promotional and non-return sales were taken.

Total weekly sales figures were found in the Store and Product matching, and the weekly average sales of each product in each store were found by averaging the weekly sales amounts on a store-product basis. To create the Pareto chart, the percentages of weekly sales averages in total sales were found and the cumulative percentages were calculated.

Products that constitute 40% of the sales volume are defined as "Fast Item", products that constitute between 40% and 80% are defined as "Medium Item", and products that constitute over 80% are defined as "Slow Item".

Store segmentation

In order to classify the stores as Fast Store, Medium Store and Slow Store, firstly non-promotional and non-return sales were taken. For store segmentation, the total weekly sales volumes were calculated only on a store basis and then the average of weekly sales volumes was calculated on a store basis.

The data was sorted from largest to smallest according to sales numbers, the percentages of sales numbers in the total data for each store and the cumulative form of these percentages were calculated and added to the data as new columns.

Finally, the stores that made up 40% of the cumulative percentage were determined as "Fast Store", the stores that made up 40% to 80% were determined as "Medium Store", and the stores that made up 20% were determined as "Slow Store".

Items that had biggest sale increase during promotions

First of all, for ease of analysis, an Is_Campaign feature was created in the form of 1-0, which shows whether there is a campaign or not. And in order to see only the effect of sales, returns were excluded by taking those with SalesQuantity > 0.

For each store-product pair, the average of campaign sales and normal sales was taken on a weekly basis, and the CampaignRatio feature was created by taking the ratio of campaign sales to normal sales.

When we look at the distribution of ratios in the chart, products with a ratio of 15 and above were considered the products with the highest sales increase during the campaigns, due to a sudden increase at and after 15.

These products codes are 60, 109, 110, 115, 156, 162, 169, 172, 174, 212, 216, 238, 239, 244, 255, 259, 264, 296 ve 301’dir.

Stores that have higher promotion reaction

The same method as above was used, maximum ratios were selected on a store basis. According to the graph below, the value of 16, where there was a sudden change, was accepted as the beginning of the Ratio of the stores most affected by the campaign. Stores with a ratio of 16 and above were considered the stores that reacted most to the campaigns.

These stores are 9, 13, 55, 73, 93, 103, 113, 156, 165, 172, 197, 203, 210, 223, 235, 236.

Observing any difference in item return rates after promotions

Return Quantity & Campaign Return Count by Weekly

For Promo1, there is a significant increase in return quantity after the campaign weeks. However, this situation was not observed in other campaign periods.

In other campaign periods, a decrease in return rates is observed after the campaign week.