PREDICTING RETAIL JEWELRY SALES

1
PREDICTING RETAIL JEWELRY SALES

• Oyvind Aassave
• Shahram Bohluli
• Rebecca Furnish
• Joneice Hamilton
• Ling-Chun Liu
• Todd Schneider

2
Brief

• Purpose of The Project
• Variables Used
• Forecasting Models Used
• Time Series
• Multiple Regression
• The Best Forecast
• Managerial Explanation

3
Purpose of Project

• Predicting Sales Amount in Year 1999
• Forecasting with the most viable variables
• Making decision about future policies
• Hiring New Staff
• Payroll
• Training
• (Passing BUSA5325 with a good grade)

4
Definition of Selected Variables

• Store Variables
• Traffic
• Payroll
• Socioeconomic Variables
• Texas unemployment rate
• US Retail (Jewelry)
• Response Variable
• Total Sale

5
Raw Data
Payroll
Traffic
Adj. Texas Unemployment
US Jewelry Retail
6
Forecasting Models

• Time Series Decomposition
• Autocorrelation
• ARIMA Box Jenkins
• Multiple Regression Models

7
Time Series - Decomposition
Trend Ratio Chart
Cycle Ratio Chart
3.5
3.5
2.8
2.8
Trend
2.0
Cycle
2.0
1.3
1.3
0.5
0.5
1995.9
1997.1
1998.3
1999.6
2000.8
1995.9
1997.1
1998.3
1999.6
2000.8
Time
Time
8
Time Series - Decomposition
Season Ratio Chart
Error Ratio Chart
3.5
3.5
2.8
2.8
Error
Season
2.0
2.0
1.3
1.3
0.5
0.5
1995.9
1997.1
1998.3
1999.6
2000.8
1995.9
1997.1
1998.3
1999.6
2000.8
Time
Time
9
Decomposition Graph
Pseudo R-Squared 0.9749707
10
Autocorrelation

Partial Autocorrelations of Total_Sales
(1,1,12,1,0)
Autocorrelations of Total_Sales (1,1,12,1,0)
1.0
1.0
0.5
0.5
Partial Autocorrelations
0.0
Autocorrelations
0.0
-0.5
-0.5
-1.0
-1.0
0.0
5.3
10.5
15.8
21.0
0.0
5.3
10.5
15.8
21.0
Time
Time
11
ARIMA Box Jenkins
Total_sales-MEAN Chart
Autocorrelations of Residuals
500000.0
1.0
375000.0
0.5
Autocorrelations
250000.0
0.0
Total_sales-MEAN
125000.0
-0.5
0.0
-1.0
1995.9
1997.1
1998.3
1999.6
0.0
8.5
17.0
25.5
34.0
Time
Lag
12
ARIMA Box Jenkins
Pseudo R-Squared 96.100066
13
The Best Time Series Model

• Decomposition sum of the residuals slightly
  smaller than that of ARIMA
• ARIMA model appears to have a better fit to the
  actual data
• No cycle is obvious from the data
• We believe the ARMIA model is better

14
Multiple Regression
15
Multiple Regression

• Equation for the best model
• -84823 13Traffic 1.08Payroll
• 62U.S. Retail Jewelry Sales
• Model with three variables seems to be the best

16
The Best Model
17
Managerial Summary

• Multiple Regression Model Proved to be The Best
  Model
• Best R-Squared
• Lowest Absolute Value of Residuals
• Noticeable Reasons For Outliers

18
THANK YOU!

• Any Questions??