Teppei Syokudo - Improving Store Performance: Evaluating Store KPIs
| Home | Product Portfolio Analysis | Improving Store Performance | Project Management | Documentation | The Team |
Overall sales, profits, and customer volume are the basic targets for food and beverage (F&B) stores. The challenge comes when attempting to achieve these targets. Sales, profits, and customer volume can be easily affected by external factors such as time period, day of the week, location, and weather. One way to achieve store targets is through store productivity. This paper looks at how store productivity can be improved in an F&B setting, through the using of various regression models.
Store productivity in F&B can be defined as the dollar amount produced per hour of work. There are three key quantitative variables that affect store productivity. The store productivity formula is as follows:
Sales – The total sales dollars earned in a time period (e.g. one hour)
Number of Customers – Number of customers served in a time period (e.g. one hour)
Number of Hours Worked – Total number of staff working during that hour
Other than quantitative variables, there are qualitative factors to consider. For example, a full-timer may generate greater productivity than a part-timer as a full-timer may feel more committed to his or her job. Also, in the presence of a manager, a staff may consciously perform better compared to when a manager is not present in the store.
To identify the factors that affect store productivity, regression analysis can be used. Regression identifies the strength and direction of relationships between dependent and independent variables. When extrapolated, it forecasts values of the dependent variable based on the strength and direction of relationship with the independent variables. Regression has been used in the fields of econometrics and law (Sykes, n.d.), in improving students’ performance (Zakhem, Khair, & Moucary, 2011), and in setting health targets (Fukada, Nakamura, & Takano, 2002).
In performing regression analysis, data-points are first plotted in a scatter plot. Scatter plots allow the user to visually identify relationships between variables. The correlation coefficient is then found. The correlation coefficient is a standardized number between -1 and 1 that describes the strength and direction of relationship between variables. A correlation coefficient of 1 or close to 1 shows that there is a strong positive relationship between the variables. A correlation coefficient of -1 or close to -1 shows that there is a strong negative relationship between the variables. A correlation coefficient of 0 or close to 0 shows that there is no or weak relationship between the variables. Using the least-squares method, the regression line and the regression equation is found. The regression equation allows us to forecast or predict the dependent variable. Regression can be split into two types: linear and non-linear. Under each type, there are simple and multiple. In this paper, we will be using the multiple linear regression model.
Multiple linear regression allows for examining how multiple independent variables are related to a dependent variable (Higgins, 2005). R (also known as the multiple correlation coefficient) is used in multiple linear regression. It represents the strength and direction of relationship amongst a combination of variables. This differs from the simple linear correlation coefficient which only compares between two variables. In knowing the strength of relationship amongst all the variables, the multiple regression formula is formed. The multiple regression formula is as follows:
Y – The value of the dependent variable
a – The intercept
b – The change in Y for each incremental change in X
X – The value of the independent variable
Care must be exercised when selecting independent variables for regression models. The variables must be truly independent of each another for the regression model to work. When independent variables are correlated with each other, the model is said to have multi-collinearity. Severe multi-collinearity can increase the variance of the coefficient estimates and make the estimates very sensitive to minor changes in the model (Farrar and Glauber, 1967). Another common problem that affects the independence of model variables is autocorrelation. The presence of significant autocorrelation in the model signals a statistical dependency between values of the same variables (Getoor, 2007). In the context of time series data, this refers to correlation between a variable’s past and future values. Autocorrelation complicates the application of statistical tests by increasing the number of dependent observations. When severe multicollinearity or autocorrelation is observed in a regression model, steps should be taken to adjust the model to reduce or remove them.
The Company
Teppei Syokudo is a Japanese F&B chain, operating under the umbrella of the famous Teppei Japanese Restaurant. It operates as a takeaway kiosk concept, and specializes in Kaisendon, a bowl of barachirashidon (rice with raw fish). It offers other dons (rice bowls), tonkatsu cutlets (fried pork), tempura (fried dishes) and sashimi salads. The takeaway kiosk mainly targets lunchtime crowds, especially upmarket businessmen and women who are looking to grab a quick, healthy, and quality lunch. There are currently 4 outlet locations - Millenia Walk, Takashimaya, Republic Plaza and Ion Orchard. For this paper, we will only be analysing data from the Millenia Walk outlet.
Each outlet has its own shop manager, cashier, packer, and kitchen staff. During peak hours, the outlets may have up to 7 staff to manage demand. Some staff work on a part-time basis. The store manager decides the staff roles for the day, hence some staff may be assigned different roles on different days. However, the job rotation usually occurs between part-timers, and rotation between cashier and packer role.
Figure 1. Sales Process
Problem Definition
One way to driving store sales is to focus on store productivity. Store productivity can be defined as the sales dollar per hour worked. However, other than quantitative measures to looking at store productivity, qualitative measures like the presence of managers, and quality of work by part-timer versus full-timer should also be considered. We come up with various hypotheses and evaluate them to identify the drivers for store productivity. Using JMP, we will evaluate the following five hypotheses:
- We can increase store productivity by hiring good cashiers who can upsell (increase sales dollar per customer) and serve customers faster (increase customer number).
- We can increase store productivity by motivating staff to work harder with the presence of managers.
- We can increase store productivity by increasing the number of full-time staff or managers present as they are more productive compared to part-time staff.
- We can increase store productivity by staffing full-timers or managers as cashiers as they are better at upselling or serving more customers than part-timers.
- We can increase store productivity by decreasing the number of staff on time periods where there is excess capacity.
