Choosing the best location based upon projected store sales

            In this paper, the data is collected from the 45 big stores and 21 small stores that would help to select future locations for store among numerous options; data was examined to find the variables that are more likely to forecast more successful store. These significant variables can also be used to construct models to forecast per square foot sales that are expected to relate to the profit of the store.

        Though, it is suggested that the concluding decision related to the selection of store location comprise real estate evaluations and cost of living to more precisely forecast the overall profit. For big stores and small stores, there are separate models developed because of the inconsistency between the trends followed by the two groups. Any stores including of than 30,000 square feet are said to be the small stores, and if the area is more than 30, 000 square feet then it called big stores.

Small Store Variables

        The selected model of small store eliminates in metropolitan classes of density, as the values of the variable fluctuate significantly from other stores and the possible options for the store not likely to include the location of this kind. There are four variables in the prognostic model that is as follow.

        XHHINC_50_74K_16TO: The household’s percentage with revenue among 50k and 74k in 16 minutes of drive-time is correlate negatively with per square foot sales. It is the most important variable with maximum effect in the entire model.

        EDUC_DOC_8TO: The overall PhD degree holders in 8 minutes’ drive-time of stores is correlated positively with per square foot sales. Scientifically, it is the second variable impactful in the entire model. Nevertheless, it is also the least importance.

        XLABOR_BLU_8TO: The blue-collar labor percentage within 8 minutes’ drive-time is a 3^rd most important variable in the model. It negatively correlates with per square foot sales, but it is also shown in the scatter plot that in the relationship, there is a large variance amount.

        LOR_AVG_1RO: The average residency length within one circular mile positively correlates with per square foot sales. Stores seem to do well in constant settings. On the model, this variable has the least impact, but not by a big change.

Big Store Variables

        For big stores, the model chosen comprises of 5 variables; one is the opening year. This variable can securely be indifferent from the model in case the individual goal is to select among the potential options of the big store location. The big stores do not seem to be considered positively affected by numerous variables. In its place, this model completely uses variables that are correlated negatively, that will help to discover a store which has the minimum possibility for negative influence from the adjacent atmosphere.

        The variable with maximum significance is the proportion of secondary degree holders within 8 minutes of the drive-time. This percentage is correlated negatively with per square foot the sales.

        Different from the small stores, big stores were more influenced by some quantity of the blue-collar labour within 8 minutes’ drive-time than the proportion. This is also said to be a 2^nd most significant variable, and it is correlated negatively with per square foot sales.

        The opening year is a 3^rd most significant variable. Before 2000 most successful stores were opened. It is expected that all options for the location might be opened in the year 2018. Consequently, this variable is not likely to support in distinguishing between the potential for profitability of the 4 locations of the big store. Though, it might offer a more precise per square foot sales forecast to compare big store with small store choice.

        The amount of women within 16 minutes’ drive-time is correlated negatively with per square foot sales. The similar is correct of men and the overall population, as these are said to be the variables that are extremely correlated. In the model, this is the 4^th most significant variable.

        In the model, the last variable is the number of competitors within 16 minutes’ drive-time. This variable is correlated negatively and offers the minimum worth to model, but not by a large quantity.

        In the above Scatterplot of the small store model, the data representation of the two variables XHHINC_50_74K_16TO and SALES_2016 can be seen. The XHHINC_50_74K_16TO is an independent variable, and the SALES_2016 is depend on the variable in this research.  It can be seen in the diagram that for the XHHINC_50_74K_16TO of 15.00 the SALES_2016 are highest and also above the 3200000. It also shows that some of the old stores have some high sales. Although the overall sales of the averages, small stores is less than 260000, that indicates that the small stores are not likely to increase their sales by a large extent over time.

Table 1. Small store model coefficients

Description of Models

Model Diagnostics for Small Stores

        The model of the small store has a standard error of estimate of 290524.66690 and Adjusted R Square of .006. This is a sensible error estimate assumed the measure of per square foot sales. Our regression model, ANOVA F-test value, is within the satisfactory range, at 1.032, with the significance of .421. This designates that this model perfectly fits in the data.

        The residuals of the model have a standard deviation of the 3171714.087 and mean of 0, with major discrepancy being a residual of 84404582074.4. The scatterplot standardized residual shows that the residuals are slightly consistently dispersed about a straight line, as anticipated.

In the above Scatterplot of the big store model, the data depiction of the two variables FPOP_16TO and SALES_2016 can be seen provided on the diagram. The FPOP_16TO represents an independent variable and SALES_2016 represents a depend on the variable in this research.  It can be seen in the diagram that for the FPOP_16TO of.00 the SALES_2016 are highest, and it is nearly reaching 4000000. It also shows that some of the old big have some high sales. Although the averages big stores overall sales are less than good enough and with high significance that indicates that big stores are more likely to increase their sales over time.

Table 2. Big store model coefficients

Model Diagnostics for Big Stores.

        The model of the small store has a standard error of estimate of 1408604.13081and Adjusted R Square of .358. This is a sensible error estimate assumed the measure of per square foot sales. Our regression model, ANOVA F-test value, is within the satisfactory range, at 7.128, with the significance of .000. This designates that this model perfectly fits in the data.

        The residuals of the model have a standard deviation of the 182509208.435and mean of 0, with major discrepancy being a residual of 7936662389358. The scatterplot standardized residual shows that the residuals are slightly consistently dispersed about a straight line, as anticipated.

Model evaluations.

        Although the model of the small store has better diagnostics model as compared with the model of a big store, the model of the big store might have more precise forecasts as compare with the model of a small store. There are 45 samples in the big store model, though the model of the small store is simply based on 21 samples. This sample size of the small model might not provide a precise depiction of the typical potential small stores behaviours, while the big store model sample size might improve the prediction correctness.

Model assumptions.

        The models that we are using are constructed to forecast per square foot sales. Though, small stores might be a suggestion of high per square foot cost, in case the prediction of the sales might be more valued as compared with the per square foot sales.

        The models are based on a hypothesis that the objective is to make the most of short-term sales. Meanwhile big stores that are older are more effective as compare with the newer big stores; it is likely that over the years big stores might uninterruptedly rise their profits. On the other hand, small stores do not seem too affected by the opening year. Consequently, big store sales in the future might reasonably increase as compared with the sales of the small store.

        Our estimates accept that small stores with concentration classes of 2 or 3 might not certainly have a similar performance as density class 4 stores.  It depends on tremendously different parameters in stores of density class 4 from data, though in the small sample size of the stores, there is a possibility that postulation is incorrect.

Conclusion on Choosing the best location based upon projected store sales

        Summing up the discussion about the furcating sales for the small and big stores on different locations it can be said that concluding decision-related to the selection of store location comprise real estate evaluations and cost of living to precisely forecast the profit. Systematically, it is the second variable impactful in the entire model, yet it is also the least importance. The big stores did not seem to be positively affected by numerous variables. The small stores are not likely to increase their sales by a large extent over time. The scatterplot consistent residual shows that residuals are consistently dispersed about the straight line, as anticipated. The small stores did not seem to effects by opening year. Consequently, big store sales in future might reasonably increase as compared with sales of the small stor