chain

MODULE 6 UNIT 1

Forecasting in the supply

© 2023 University of Cape Town

All Rights Reserved
Table of contents
1. Introduction 3
2. The importance of forecasting and demand management 3
3. Qualitative forecasting techniques 4
3.1 Panel approach 4
3.2 Delphi method 5
3.3 Scenario planning 5
3.4 Market research 5
4. Quantitative forecasting techniques 5
4.1 Time sales analysis 6
4.1.1 Moving-average forecasting 6
4.1.2 Exponential smoothing 6
4.2 Naive forecasting 7
4.3 Causal modelling 7
5. Measuring forecast error 8
6. Conclusion 8
7. References 9

© 2023 University of Cape Town

All Rights Reserved

Tel: +27 21 447 7565 | Fax: +27 21 447 8344

Website: www.getsmarter.com | Email: [email protected]
Page 2 of 9
 Learning outcomes:

LO1: Identify the different forecasting methods and their applicable contexts.

LO2: Determine the differences between various quantitative and qualitative forecasting
techniques.

LO3: Interpret the role that forecasting error plays in ensuring an operation’s success.

1. Introduction
Forecasting is the process of making a prediction about a future event while considering the
performance or trends of previous events (Kolassa & Siemsen, 2016). In a supply chain
context, forecasting is used to predict how much supply will be needed, where, and when.
This can be a challenging task to perform accurately, but by understanding the differences
between qualitative and quantitative forecasting techniques, as well as how and when to use
them appropriately, the task becomes more manageable.

Forecasts can be created through various methods, such as statistical models and human
judgement. Statistical models often rely on algorithms and convert data into forecasts through
software or manual spreadsheets. Human judgement methods rely on intuition alongside
statistical data to inform decisions. Both are used to some degree in all forecasting techniques
(Kolassa & Siemsen, 2016).

This set of notes will guide you through the practice of forecasting, the methods involved, and
how to account for error.

2. The importance of forecasting and demand

management
Every leader should have an eye on the future growth of their business, which means that
their decisions are often based on forecasts. These forecasts suggest what they can expect to
happen in the future to support accurate and efficient planning in the present. However, as
no forecast is perfect, confronting, quantifying, and managing uncertainty is an important part
of the planning process (Kolassa & Siemsen, 2016).

As with data-capturing techniques, forecasting consists of quantitative and qualitative

strategies. A qualitative approach involves making judgements about the future using
opinions or experiences to guide decision-making. This approach is often used when
evaluating what competitors are experiencing, based on management’s viewpoint of their

© 2023 University of Cape Town

All Rights Reserved

Tel: +27 21 447 7565 | Fax: +27 21 447 8344

Website: www.getsmarter.com | Email: [email protected]
Page 3 of 9
situation. A quantitative approach focuses on hard data, using data points to outline
behaviour patterns over a period of time. This approach can be complex (Slack et al., 2017).

As uncertain as forecasts can appear, they are an important part of operations, particularly in
the supply chain. Forecasts form the foundation of all decisions and can be used to reduce
costs and waste and improve responsiveness and customer service (Sanders, 2016). In the
following sections, you will learn about qualitative and quantitative forecasting techniques in
more detail.

3. Qualitative forecasting techniques

Consider the case of Strictly Cotton, used in previous modules. When the business needs to
decide what the next season’s in-demand pattern and colour will be, it must do both
qualitative and quantitative forecasting to develop an accurate prediction.

Effective qualitative forecasting techniques require an understanding of past performance of

the brand, as well as “expert” guesses. For Strictly Cotton, these expert guesses might use the
knowledge of fashion writers, professional buyers, and shop assistants who deal with
customer opinions and desires every day. These opinions, as well as management’s
judgements, are collected and analysed to help leaders to make a decision.

Imagine, for instance, that a customer of Strictly Cotton were asked to forecast brand trends.
They would likely factor in personal preference and rely heavily on qualitative techniques.
However, this alone probably would not result in an accurate forecast. To get to a more
accurate answer, the customer would need to consider other factors, including what the last
few seasons had in common or what they have seen in the media recently. They may even ask
a few friends.

Different industries see various types of qualitative forecasting, covered in the following
sections.

3.1 Panel approach

In a panel approach, a group of experts in an organisation, such as the sales staff or managers,
meet and develop a forecast collectively. There is no formal structure to this method, and it
is based on open discussion among the group members. In this method, a leader incorporates
the expertise of those who are already working in sales and may already know of a trend
coming – for example, more customers may be asking for red jackets or shops may be running
out of boot-cut jeans faster than usual. This up-to-date information can be added to the
discussion to get a clearer idea of what is happening on the ground.

The downside of this method is that it is vulnerable to bias if one loud voice dominates the
discussion or there is unequal participation (Sanders, 2016).

© 2023 University of Cape Town

All Rights Reserved

Tel: +27 21 447 7565 | Fax: +27 21 447 8344

Website: www.getsmarter.com | Email: [email protected]
Page 4 of 9
3.2 Delphi method
The Delphi method works with an anonymous survey to gather experts’ opinions and guesses.
The survey results are summarised, analysed, and returned to the experts who answered
them. This process allows experts to re-evaluate their answers based on what other expert
opinions reveal. The process is repeated until there is either consensus or the prediction has
been refined to the closest option available.

Challenges of this method include selecting the correct experts, developing the most accurate
and unbiased questions, and attempting to account for bias in the subjects (Slack et al., 2017).

3.3 Scenario planning

Scenario planning is an adjusted panel approach focused on the greater uncertainty of long-
range forecasting. The panel develops one or more likely future scenarios and analyses their
related risks. Scenario planning does not focus on finding consensus; instead, it aims to put
plans in place that account for all (or at least most) of the scenarios to avoid the least-desired
outcome and push ahead with the desired one (Haksever & Render, 2017).

3.4 Market research

Market research uses surveys and interviews to gather information from consumers instead
of experts. Information on consumer preferences and queries are gathered from as wide a
range of customer demographics as possible. Leaders can use these findings to ensure
consumers get what they want.

Market research is a good determinant of customer needs and preferences, but it can be quite
time-consuming to collect and sort data from a large sample size. Market research also falls
short when there are not enough well-rounded questions that fit all customers and account
for all possible attitudes. For these reasons, it is important to consider hiring an outside
market-research firm to conduct the study to ensure surveys are reliable and an appropriate
variety of consumers are surveyed (Sanders, 2016).

4. Quantitative forecasting techniques

In quantitative approaches, patterns and behaviours are analysed and contextualised
according to cause-and-effect relationships. These can be based on known facts concerning
certain events or actions leading to certain predictable outcomes. Alternatively, some work
should be done to draw conclusions on which preceding events or actions contributed – and
by how much – to the outcomes.

For example, Black Friday marketing and sales highlight a cause-and-effect relationship. When
marketing for Black Friday, retailers advertise specials ahead of time, allowing pre-screening

© 2023 University of Cape Town

All Rights Reserved

Tel: +27 21 447 7565 | Fax: +27 21 447 8344

Website: www.getsmarter.com | Email: [email protected]
Page 5 of 9
and online buying by registered shoppers. This can cause a spike in website subscriptions and
an increase in impulse purchases. To accurately forecast Black Friday supply and marketing, it
is possible to view the demands of Black Friday sales in previous years to come up with a
strategy for the next.

Quantitative forecasting methods can use one of several approaches to analyse data, explored
in the following sections.

4.1 Time sales analysis

Time sales analysis considers a single phenomenon over time and studies the patterns of past
consumer behaviours, and what may have caused any changes, to pre-empt future behaviour
(Slack & Brandon-Jones, 2019). Simply put, it tracks whether sales trends are rising, falling, or
staying consistent, and projects future sales based on the assumption that this trend will
continue. A limitation of this method is that it can only take past patterns and slower moving
trends into consideration and account for rapid changes or disruptive factors.

Time series analysis is further broken down into two approaches: moving-average forecasting
and exponential smoothing.

4.1.1 Moving-average forecasting

Moving-average forecasting takes the actual demand figures over a predetermined period –
for example, the previous four weeks – to calculate average demand over time. This average
is then set as the forecast for the next period (Slack et al., 2017).

Example:

A bed store decides its weekly orders based on time series analysis using a moving-average
forecast. In the previous four weeks, the store sold 10 beds in the first week, 20 beds in
the second, 30 beds in the third, and 40 beds in the last week.

The total for the four weeks is 100 beds and the average is 25 beds a week, which is used
as the forecast for the fifth week in order to avoid extreme stock shortages or excesses.

The main drawback of using moving-average forecasting is its reliance on averages. In reality,
it is quite unlikely that any week will perfectly align with the average. It also does not take any
other points of potentially confounding data into consideration.

4.1.2 Exponential smoothing

Exponential smoothing attempts to correct for the shortcomings of moving-average
forecasting by taking into consideration the actual demand of the current period (which would
be Week 4 in the previous example) and what the forecast was for that period. It incorporates

© 2023 University of Cape Town

All Rights Reserved

Tel: +27 21 447 7565 | Fax: +27 21 447 8344

Website: www.getsmarter.com | Email: [email protected]
Page 6 of 9
the previous week’s data and forecasting for the purposes of comparison, and keeps in mind
the difference between the current demand and forecast to allow more stability between
forecasts from week to week (Slack & Brandon-Jones, 2019).

Example:

If the demand for Week Four was 40 beds but the forecast had been 20, the difference of
20 is used to adjust the average. Therefore, the new forecast for Week 5 is 45 beds.

Exponential smoothing is an effective and simple forecasting technique that includes both
trend and seasonality components. In most cases, the models used for this form of forecasting
are somewhat more complex than the above example. Luckily, software assists with
exponential smoothing and there are over 30 models that can be used depending on the data
samples (Kolassa & Siemsen, 2016).

4.2 Naive forecasting

Also known as demand chasing, naive forecasting uses the most recently observed demand
to predict the future. The term “naive” implies that it is simplistic (Kolassa & Siemsen, 2016).
Naive forecasting simply uses the previous periods’ actual demand as the next period’s
forecast. This method is best used as a benchmark of performance rather than an actionable
forecast (Sanders, 2016).

Example:

The bed company sold 40 beds in the last week, so it predicts that it will need to have at
least 40 on hand for the following week. It also uses this number of 40 beds sold as its
performance goal; any more or less sold will be used as a motivation for some intervention
aimed at improving dropped numbers.

4.3 Causal modelling

Causal modelling analyses relationships between datapoints and their effect on performance
(Haksever & Render, 2017). This method uses multiple complex techniques for predicting
sales. No single factor, but rather multiple variables are tracked and evaluated, as causal
modelling assumes that relationships between a number of factors affect performance and so
should also impact planning (Slack & Brandon-Jones, 2019).

© 2023 University of Cape Town

All Rights Reserved

Tel: +27 21 447 7565 | Fax: +27 21 447 8344

Website: www.getsmarter.com | Email: [email protected]
Page 7 of 9
 Example:

In this case, the bed company would also take into consideration its marketing strategy –
perhaps it was selling more beds based on a discount or new ad campaign. It could include
the percentage of purchases made online or in store. It could factor in whether a certain
brand sold more and in which areas it had adjusted sales figures before forecasting the
next week. All these figures can influence the next period’s forecast.

5. Measuring forecast error

The common denominator in all forecasting is imperfection. It is extremely rare for any
prediction to be exactly correct. Simply put, forecasting error is the difference between what
was predicted and what occurs. To help account for inevitable errors, forecasting error
measurements can be used to get an idea of how well the organisation is doing in forecasts
and whether adjustments are needed (Sanders, 2016).

Forecasting error cannot be avoided, but what can be managed is the size of the mistakes in
the results. There are a few ways to measure forecast error:

• Mean absolute deviation (MAD): This is a calculation of the average of the absolute
errors, i.e. the total difference between the predicted and actual numbers for certain
periods.

• Mean percentage error (MPE): This is a calculation that measures the average or
mean of errors as a percentage.

Both measures try to capture an error trend in predictions that can be incorporated into new
results in an attempt to swing final predictions towards a likelier reality. For example, if on
average forecasts are off by a MAD of 30 or an MPE of 5%, this error can be accounted for by
adding or subtracting that amount from the forecast.

Most organisations use software to generate these measurements, but it remains important
for the supply chain manager to understand how to interpret and use the results to predict
supply chain needs that relate to placing future orders, reserving storage space, and making
data-driven decisions for running the supply chain.

6. Conclusion
Without some estimate of the future and what it will require, a leader cannot plan
successfully. Forecasting involves predicting what to expect and what to aim for. They are not
targets, plans, or budgets, but should be used to help develop them.

© 2023 University of Cape Town

All Rights Reserved

Tel: +27 21 447 7565 | Fax: +27 21 447 8344

Website: www.getsmarter.com | Email: [email protected]
Page 8 of 9
Understanding the available data is the first step to a good forecast. To forecast in the most
accurate manner, it is best to bring together a combination of qualitative and quantitative
techniques that consider hard data and customer sentiment and attitudes, and not leave any
potential influencing factor unaccounted for.

Reflect on your learning:

What are your biggest takeaways from this set of notes? Record your reflections in your
journal. Remember that the course-wide journal can be downloaded in Module 1.

7. References
Haksever, C. & Render, B. 2017. Service and operations management. Singapore: World
Scientific Publishing.

Kolassa, S. & Siemsen, E. 2016. Demand forecasting for managers. New York: Business
Expert Press.

Sanders, N. 2016. Forecasting fundamentals. New York: Business Expert Press.

Slack, N. & Brandon-Jones, A. 2021. Operations management. 9th ed. Harlow, UK: Pearson.

Slack, N., Brandon-Jones, A., Johnston, R., Singh, H. & Phihlela, K. 2017. Operations
management: global and Southern African perspectives. 3rd ed. Cape Town: Pearson.

© 2023 University of Cape Town

All Rights Reserved

Tel: +27 21 447 7565 | Fax: +27 21 447 8344

Website: www.getsmarter.com | Email: [email protected]
Page 9 of 9