ALDI Melbourne: Cost-Effective Staff Scheduling During Covid-19 Crisis

This report explores cost-effective staff scheduling strategies for ALDI Melbourne during the Covid-19 crisis. It begins with an introduction to ALDI's presence in Australia and its response to the pandemic, followed by a literature review of relevant articles and journals identified through Google Scholar using keywords like 'Staff scheduling' and 'Covid-19 Pandemic and Retail Staff'. The methodology section details data collection challenges, including acquiring staff data from ALDI and using Google Analytics for customer foot traffic estimates. Text data mining was performed on Google reviews to identify customer concerns, particularly regarding wait times and queues. Predictive analysis, including regression planning, is proposed to forecast staffing requirements based on historical data and customer demand. The report uses descriptive, predictive, and prescriptive analysis techniques to provide actionable recommendations for optimizing ALDI's staff scheduling practices during the ongoing pandemic.

Introduction
ALDI is one of the most popular supermarket chains available in Australia. Despite
being a German based discounted supermarket chain, it has grown with immense popularity
across the Australian population. As mentioned by Mortimer & Grimmer (2021), ALDI
started its operations in Australia back in 2001 with two stores, one of them in Sydney’s
Marrickville and the other in Bankstown Airport. The company essentially started
functioning back in 1913 but since the commencement of their operations in Australia, it has
expanded its operations across 570 stores spreading in six states and territories eventually
employing more than 13,500 people (Aldi.com.au, 2021). ALDI’s success in Australia is
eventually attributed to their ability to understand the gap in the Australian grocery market
which was the lack of discounted supermarket chains.
ALDI has been known to address the Covid-19 crisis with a suitable plan that would
permit them to achieve their daily objectives while maintaining good customer experience.
As mentioned by Redman (2021), ALDI not only encouraged its employees to acquire
vaccination for incentives but provided scheduling flexibility to its salaried employees.
Furthermore, the company also went on hiring spree to manage the traffic in its Melbourne
store and their other outlets on a global scale especially when the traffic was continuing to
increase (Uschamberfoundation.org, 2021). Since the onset of the covid-19 pandemic, the
ALDI management and the employees collectively have been working to provide necessary
services to the customers without any interruption.
The aim of this report is to focus on the staff scheduling prospects of ALDI
Melbourne in a much cost-effective way until the covid-19 crisis has been resolved. The
structure of this report has been categorised into five different sections consisting of an
introduction, literature review, methodology, analysis of the acquired data using analytical

techniques, ending with conclusion and recommendations. The analytical techniques used in
this report would consist of descriptive analysis, predictive analysis and prescriptive
analysis.
Literature Review
The preferred mode for identifying key articles and journals associated with the
subject matter was the use of ‘Google Scholar’ database because the accessibility to wide
range of journal articles and research papers from different years get much easier. The use of
certain keyword was prompted to ensure that accurate results were identified by the end of
the search such as ‘Staff scheduling’, ‘Retail Market of Australia’, ‘Covid-19 Pandemic and
Retail Staff’, and more. The use of such keywords make it easier for the learner to identify
the exact articles that would align with the perspective of the study and would enable in
achieving the purpose.
Some of the companies around the world focused on establishing a good relationship
with the customers to ensure that they can genuinely address customer needs and navigate
their functionality surrounding it accordingly. According to Dore et al. (2020), the idea of
staff-sharing plan used by companies such as ALDI Germany and McDonalds Germany
showcases how collective participation makes it easier to handle crisis. It also paves way for
securing the employment for the furloughed employees or the ones working temporarily.
McDonald’s temporary workers were redeployed at ALDI stores to ensure the effectiveness
of the supply chain and proper deliveries to its customers. Dore et al. (2020), explored the
idea that not only this was a good employment opportunity for unemployed individuals but an
effective way to handle understaffing problem mainly because the staffs were affected by
covid-19. Even though this article shares a solution of staff-sharing to deal with understaffing
issues and scheduling conflicts but this strategy would only be beneficial for short-term

effectiveness and would not aid the organisation in long-term. This strategy can still be used
as an extreme measure for a company which is facing severe issues in terms of staffing
schedules.
The aspect of staff scheduling was a major concern during the covid-19 pandemic for
majority of the supermarket chains functioning in Australia, which prevented them from
assuring the efficiency of the workforce management. Guerriero & Guido (2021), mentions
that staff scheduling problems can be easily addressed when the level of employee
satisfaction is considered by an organisation and with the use of novel optimization models, it
gets much easier to provide flexibility to the staff. The novel optimization model has to be
focused on the different working times along with employee availability and preferences.
Eventually, despite flexibility added to schedules, the purpose should be to achieve maximum
employee satisfaction and the demand requirements of the day. However, Guerriero & Guido
(2021), fails to organise every objective function of an organisation into one model because
the integration of six different optimization models would be extremely chaotic and difficult
to handle. Thus, the preference is laid on bi-objective optimization formulation and novel
optimization, among which the former considers employee availability and potential demand
requirement.
The scheduling challenges also emerge when the consumers are not available in usual
quantity that the stores are used to regularly. Due to the occurrence of the covid-19 pandemic,
many of the consumers around the world and in Australia preferred the use of online
purchases instead of going to the store to purchase necessary items. The shopping frequency
of the consumers reduced from 2-3 times a week to once a week where the flexible daily
schedules of the employees at the organisation made them feel overworked and resulted in an
unstructured schedule (Wang et al., 2020). This article has successfully highlighted how
changes in consumer behaviour can also prompt change in store’s employee management in

the long-run. It is one of the biggest strengths of the article and that consumer shopping
frequency, timing, duration, expenditure and even their switch to online purchases are
reflected. However, Wang et al. (2020), did not attempt to explore in depth how it could
affect the scheduling conundrum of working employees of a store. Thus, even though this
article points out a different element to consider, it is not entirely researched in an in-depth
manner.
The staff scheduling specifically emerged during the covid-19 pandemic partly due to
the guidelines mandated by CDC to assure hygiene and avoid extensive gatherings within the
store. According to Nakat & Bou-Mitri (2021), the companies were encouraged to revise
their sick leave policies, accordingly update the cleaning schedule and minimize contact
among workers as much as possible through shift staggering. The changes in work hours and
the implementation of additional short breaks for all the staff made it mandatory for
companies like ALDI to go on hiring spree especially at the initial stage when consumers
were busy hoarding food items for their consumption. Considering the health factors
affecting staff scheduling during covid-19 is one of the key strengths of this article but the
management decisions undertaken by companies when employees contracted the virus have
not been researched properly. This can be prompt as a weakness of this article.
The employee sharing model has been effectively successful across different sectors
in the times of crisis because technological advancements have severely transformed the
industry making it mandatory for the ones operating in the retail sector to make relevant
changes. De la Mora Velasco, Huang & Haney (2021), agrees with Dore et al. (2020), and
commences the use of employee sharing model (ESM) for the hospitality industry, which is
easily applicable for the supermarket chain as well. De la Mora Velasco, Huang & Haney
(2021), highlights that the use of an ESM would be effective in addressing unemployment,
enhance the responsiveness of the employees along with sufficient profitability. The

employees can work for multiple workers depending on the schedules they can collectively
work on. One of the key strengths of this model is that employee preference would be highly
prioritised and a relationship would be established on a contractual basis. However, one of
the key weaknesses of this article is that it does not comply with the traditional employment
schemes and how to align the arrangement accordingly.
Zucchi, Iori & Subramanian (2021), agrees with Guerriero & Guido (2021), and
propose the use of bi-objectification model, which not only provides with a comprehensive
idea of how the employees would be hired and arrange as per schedule but is also focused on
the assuring fair work-shift distribution. However, the expertise of the employees would have
to be considered accordingly as well. Without that it would be difficult for companies like
ALDI to progress in the market while dealing with a crisis. Zucchi, Iori & Subramanian
(2021), highlights that the company may focus on adding a skill matrix to delegate work
hours depending on the skills and expertise of an individual. However, the key weakness of
including this matrix would be that some employees might remain overworked.
The identification of strengths and weaknesses of the articles suggested that the use of
employee sharing model and the bi-objectification model would be beneficial for ALDI to
arrange flexible staff scheduling while dealing with covid-19 pandemic. Furthermore, to align
with CDC guidelines and changing consumer behaviour getting into contractual based
temporary arrangements with employees with focus on their terms and conditions would be
beneficial for the company.
Methodology and Implementation
Descriptive Analysis
Data Collection

In line with concerns over how COVID-19 impacts supermarkets, data on the number
of staff were sought from the ALDI CBD store. Unfortunately, the Store Manager stated that
any information could not be given regarding staffing schedules and shift timing. Instead, a
sample schedule was sought from an equally sized Coles store with similar opening hours.
Some measurements of customer data were taken at the ALDI store, though with lockdown
restrictions, collecting a true sample was difficult. Instead, Google analytics of live “foot-
traffic” were used as a scale (Figure X). The small sample of observations of actual number
of customers were attributed to this scale to find a more representative average which was
then multiplied across the scale, providing an indication of the number of customers at any
time point across the week (calculations can be seen in Descriptive.xlsx).
Figure X. Example of Google “foot-traffic” for ALDI. The horizontal lines represent the
scale multiplier used (from 0 to 3).
Text Data Mining. Key data for the following analysis revolves around how ALDI
operates and if there are any shortfalls in their employee scheduling practices. Text data
mining was undertaken by collecting all 2,337 reviews from Google and excluding those
without any comments (i.e., those which were a rating only). The resulting pool of 914
reviews provided initial information that the store was, overall, highly rated with an average
of 4.2 stars. Target-words such as “bad” and “queue” along with close associations such as

“poor”, “checkout”, and “line” were employed across all reviews. The frequency of these
terms was collated and for more accurate comparison, were equated to proportional
frequencies based on the stars given for each review. As such, all 1-star reviews are treated as
a single document, all 2-star reviews are treated as one document, and so on. Using k-means
clustering, two distinct clusters were evidenced (Figure X): one which comprised low
proportional frequencies of the words “bad” and “queue” (3-, 4-, and 5-star reviews) and one
with high proportional frequencies composed of 1- and 2-star reviews.
Figure X. k-means clustering analysis of proportional frequency for target-words “bad” and
“queue” as distinguished by number of stars given.
This analysis revealed that poor reviews are typically concerned with wait times and
slow queues. Further predictive and prescriptive analysis is used to counteract this by

providing employee shift assignment which caters to busier periods, limiting pressure on the
checkout lanes and minimising slow queues.
Predictive Analysis
Regression Planning
To maintain store efficiency and keep up with the demand of shoppers during busier
periods, employee requirements need to be calculated. Specifically, a regression equation can
be used to predict the number of staff required for any time-point by using historic staff
numbers and customer data. Day and hourly time slots were converted to dummy variables,
with base groups of Monday and 8 – 9am respectively. The general equation is formulated:
𝑌 = 𝛽0 + 𝛽1 𝑋ij + 𝛽2 𝐷2 + 𝛽3 𝐷3 + 𝛽4 𝐷4 + 𝛽5 𝐷5 + 𝛽6 𝐷6 + 𝛽7 𝐷7 + 𝛽8 𝑇2
+ 𝛽9 𝑇3
+ 𝛽10𝑇4 + 𝛽11 𝑇5 + 𝛽12 𝑇6 + 𝛽13 𝑇7 + 𝛽14 𝑇8 + 𝛽15 𝑇9 + 𝛽16 𝑇10 + 𝛽17
𝑇11
+ 𝛽18 𝑇12 + 𝛽19 𝑇13 + 𝛽20 𝑇14 + 𝛽21 𝑇15 + 𝛽22 𝑇16 + 𝜇
Y = number of staff (independent variable)
Xij = number of customers on day i at time j (dependent variable)
Di = day of the week (dummy variable)
D2 = Tuesday
D3 = Wednesday
D4 = Thursday
D5 = Friday
D6 = Saturday
D7 = Sunday
Tj = time of day (dummy variable)
T2 = 9 – 10am
T3 = 10 – 11am
T4 = 11am – 12pm
T5 = 12 – 1pm
T6 = 1 – 2pm
T7 = 2 – 3pm
T8 = 3 – 4pm
T9 = 4 – 5pm
T10 = 5 – 6pm

T11 = 6 – 7pm T12 = 7 – 8pm
μ = error term
Multicollinearity. To ensure there were no concerns over multicollinearity, a
correlation matrix was produced (Figure X). A simple rule-of-thumb is for values above 0.7
or below -0.7 to be cause for concern, as this generally indicates that a secondary correlation
occurs between two or more of the independent variables. By design, the dummy variables
will not have any multicollinearity effects within each other, so the focus is on their
correlation with the “Customers” column. As none of our dummy variables indicate an issue
here, the regression equation remains viable.
Figure X. Correlation matrix, focussed mainly on values for column two: “Customers”
Regression Model
The software R Studio was used to estimate the coefficients for the planned regression
equation. The code used is quite simple, and the output is saved to a text file and then
converted into Table 1. For ease, the times were converted from hourly brackets to Tn. The
key line to generate the model is given below:
model <- lm(Staff ~ Customers + Tuesday + Wednesday +
Thursday + Friday + Saturday + Sunday + T2 + T3 + T4 +

T5 + T6 + T7 + T8 + T9 + T10 + T11 + T12, data =
regression_data)
The ordinary least squares method is employed to generate the coefficient values
which are substituted for the 𝛽n placeholders from earlier.
Table 1.
Dependent variable: Number of staff
Note: *p < 0.1; **p < 0.05; ***p < 0.01
Independent variables Coefficient Std. Error
Constant 2.921*** 0.236
Number of customers 0.012*** 0.002
Tuesday 0.461** 0.208
Wednesday -0.039 0.210
Thursday -0.061 0.212
Friday -0.232 0.208
Saturday 0.360* 0.262
Sunday 0.126 0.241
9-10am 0.134 0.284
10-11am 0.274 0.317
11am-12pm 0.129 0.364
12-1pm 0.106 0.394
1-2pm 0.639** 0.413
2-3pm 0.686** 0.408
3-4pm 0.808* 0.395
4-5pm 1.036*** 0.416
5-6pm 0.506 0.491
6-7pm -0.055 0.474

7-8pm -1.166*** 0.340
Observations 82
R2 0.896
Adjusted R2 0.868
Residual standard error 0.509 (df = 65)
F statistic 31.267*** (df = 18; 65)
Thus, the regression equation can be rewritten as:
Staff required=2.2921+0.012 ( Number of customers ) +0.461 ( Tuesday ) −0.039 ( Wednesday ) −0.061 ( Thur sday
Interpretation
For brevity, the following analyses only those parameters which are statistically
significant (p < .05).
𝛽0 = constant; minimum number of employees required at base groups (8 – 9am on Monday)
𝛽1 = each additional customer increases the requirement for workers by 0.012; approx. 83
customers equate to needing one more worker
𝛽2 = it is the time duration which is required on Tuesday
𝛽3 = it is the time duration which is required on Wednesday
𝛽3 = it is the time duration which is required on Thursday
𝛽3 = it is the time duration which is required on Friday
Implementation
The extrapolated customer data based on Google foot-traffic proportions is then used
in the regression model to predict future demand for service staff. For example, projected
demand for staff on Saturday at 12pm is calculated as:

ŷ = 2.921 + 0.012(328) + 0.461(0) – 0.039(0) – 0.061(0) – 0.232(0) +
0.360(1)
+ 0.126(0) + 0.134(0) + 0.274(0) + 0.129(0) + 0.106(1) +
0.639(0)
+ 0.686(0) + 0.808(0) + 1.036(0) + 0.506(0) – 0.055(0) – 1.166(0)ŷ = 2.921 + 3.936 + 0.360 + 0.106 = 6.693 🡪 rounded to 7 staff
The predicted values were calculated for each hour of each day. The results indicate
the required staff for any time slot during the week’s opening hours and form the RHS
constraints (minimum staff) in the prescriptive analysis.
Prescriptive Analysis
Linear Programming Planning
Shift scheduling. The shift schedule was formulated from a realistic assumption of
how ALDI’s shifts may be formatted, shown in Figure X. To demonstrate both job roles,
casual and full-time (FT) employees have been used, though part-time staff were excluded for
simplicity. Overtime is not included in this model and given the opening hours (8am – 8pm),
there is always sufficient break between shifts on any two consecutive days. FT employees
receive a 1-hour paid break during their 8-hour shift; casuals receive no break across a 4-hour
shift.