Executive Summary

After analyzing 15 locations of Nick’s grocery store, we determined that the semi-log model gave us the most useful information when compared to other models. Based on our results, we decided to keep the model 10. When prices are nearly identical, we recommend adjusting prices to have ending with 9, instead of only focusing on the 1 cent price increase. Products are doing well with an estimation of 43.46% increase after two years, but researchers found that the other three seasons were nearly 20% greater than the winter season, which, in context, is to be expected as orange juice tends to be a warm-weather beverage. Product cost should be provided for further price determination.

Introduction and Background

Nick’s is a grocery store located in the Midwest with 15 locations. The regionally based company needs help with standardized pricing in their stores. Our team has been assigned to: Tropicana 64 oz., the largest selling product in the orange juice category. The dataset researchers used are sales between January 2009 - December 2010 for the product. Our team will focus on determining price setting strategy, storage location with limited supply and present the product sold quantity based on each season to adjust our stock quantity.

Data and Methodology

The dataset this project uses is weekly sales volume of Tropicana 64 ounces, processed including columns of store, week, quant, price, and deal. We are using the semi-log model to do the analysis. Under semi-log, the slope coefficient will measure the relative change in Y-axis for a given change in the value of the explanatory variable. We are using SAS for the analytic software, implementing project_1_addtional_SAS code. A few of the basic descriptive statistics ^[1] we calculated were mean, median, maxima, minima, standard deviation. Mean represents the average number ounces of juice that is sold in each store per week. Maximum and minimum each represents highest and lowest number of ounces of juice that is sold per store per week respectively. Median represents the “middle” number of ounces sold per week, taken by ordering all values of orange juice sold and choosing the middle number. Standard deviation depicts how dispersed our data are.

Key Findings

According to the results, we will keep the model 10 and discard other models. First, the Adj R-Sq of other models' are less than model 10, indicating that other models' relationship between the statistical model and variables are not as strong as model 10. Second, adj R-Sq ^[2] of Model 10 is comparatively higher than other models (the value is 0.474, which is very close to 0.5) We noticed that other models do not include all the variables that may affect the product sales. On the statistical side, most of the PR values of the variables on model 10 are lower than 0.0001. These variables are quite consistent with the model. With all these perspectives ^[3], the other models would not be better choices compared to model 10.

After estimating data presented through SAS with relevant factors included ^[4], by distributing different weeks into a dummy variable “qrt 1-3” our research indicates the product the company is selling might be affected by different seasons. Qrt 1, with the parameter of 0.21152, compared with qrt 4; refers to winter season, this means spring season had a sell of

e^0.21152 = 1.2356 which means expected spring sales is 23.56% greater than winter season. This is statistically significant with the P-value less than 0.001, compared to the threshold of 0.05. Qrt2 and Qrt3 had a sell of 119.22% and 115.75% and both were statistically significant. Each store dummy variable represents individual store sales compared to store number code 137. With negative parameters, meaning they all had a selling less than store code 137. As of store1, with parameter of -0.84167, their sell quantity = 1/e^0.84167 = 0.4309 = 43.09% of store code 137. With T- value less than -2, and P- Value less than 0.01 indicates it is statistically significant.

Research also suggests setting prices to end with the number 9. A price of $4.89 had a selling quantity increase of e^0.18746 = 1.20618 = 20.62% sell increase compared with market price that does not end in 9. This is statistically meaningful to consider, and the product is in a good shape each week, increasing the selling quantity by 0.34% than previous week. At the end of second year our selling quantity would increase to e^00347^104 = 1.43459 * 100 = 143.46% compared to the first week.

Conclusions and Recommendations

As shown above, we consider the semi-log model including variables of all stores, price, deal, price ending with 9, week and quarter is a better fit among other models. The reason we prefer this model is because it reflects the most factors that we consider meaningful. Recommendations are provided based on various parameters as follows:

Location: Defining all negative parameters though store 1-14 The probable reasons are their location, interior decoration, demographic nearby etc. Researchers recommend, when quantity of products is limited, more products should be allocated to store15.

Price: Overall profit follows function . To get the maximum profit, we make the derivative of the formula equals 0. After algebraic manipulation ^[5], we yield . The optimal price is cost(2.57) minus the reciprocal for price coefficient (-1.54). dollars

End9: Product price ending with 9 has a positive coefficient. It means end9 strategy is useful in boosting the overall sales. When prices are nearly identical, we recommend adjusting prices to have ending with 9, instead of only focusing on the 1 cent price increase.

Quarter: Quart1, quart2 and quart 3 all have positive coefficients, quarter4 has the least quantity expectation compared to other 3. Quarter 4 is considered to have the lowest temperature in the northern hemisphere. Tropicana is stored in a refrigerated environment and is often consumed cold. So, our researcher recommends product inventory should be reduced accordingly in quarter4.

The study has several limitations that are worth mentioning. Optimal price is yielded based on other parameters staying unchanged, while pricing could affect deal and other parameters. So, interactions with others should be studied further. Local demographic, location within the store, package size are also meaningful factors to be considered for future research.