22 Resources
This chapter summarizes additional resources for you to learn more about forecasting. We categorize these resources by grouping them into different sections for non-technical, somewhat technical, and technical material. The present book would be in the “somewhat technical” section. The resources in the “technical” section require more technical background knowledge, typically in terms of statistics. We also point to forecasting organizations, events, datasets, and online resources.
22.1 Non-technical material
Gilliland (2010) focuses on the process aspects of forecasting. In our opinion, its most valuable parts are two chapters on worst practices in forecasting and one chapter on Forecast Value Added analysis, a framework that aims to capture how much value a forecast (or specific steps in a forecasting process) adds in a business sense (see Section 16.3).
Goodwin (2017) summarizes what Professor Paul Goodwin at the University of Bath has learned about forecasting in his decades-long career studying the topic. It paints a broad canvas and provides insights into many different areas of forecasting.
Tetlock and Gardner (2015) describe a sociological experiment on forecasting. Are there people who can forecast very well? If so, how can we help them become even better? The intellectual origins of this book lie in the “Wisdom of Crowds” phenomenon, i.e., the argument that forecasts should rely on average predictions from a group rather than the opinions of individual experts. Somewhat ironically, this book concludes that so-called “super forecasters” exist, i.e., people who are remarkably accurate in their predictions. The book focuses less on demand than geopolitical forecasting (the CIA enabled this experiment). While there is no immediate applicability to better demand forecasting, the book is insightful and a joy to read.
Vandeput (2023) makes very similar points to us in his collection of best practices in demand forecasting. His description of what to look out for and what pitfalls to avoid is informed by many years of practice and experience “in the trenches” of supply chain forecasting. This book makes for an excellent companion to the one you have just read.
Foresight: The International Journal of Applied Forecasting (https://forecasters.org/foresight/) is a quarterly publication of the International Institute of Forecasters aimed at practitioners. Articles are written for and by practitioners. Academics occasionally contribute, but they write their articles with a practitioner audience in mind. The journal frequently publishes features, i.e., a longer article on a specific topic and commentaries by other authors. It also publishes books compiling articles on a common theme, e.g., Gilliland et al. (2015) and Gilliland et al. (2021). (Full disclosure: one of the present authors, SK, is a Deputy Editor at Foresight.)
22.2 Somewhat technical material
Kolassa and Siemsen (2016) is an earlier book by two of the authors of the book you are currently reading. We based our present work on it but expanded it significantly.
Hyndman and Athanasopoulos (2021) provide a more technical introduction to forecasting. The authors are two of the foremost academic forecasting experts of the day. If you want to know how to derive ARIMA orders from an ACF/PACF plot or how optimal hierarchical reconciliation of forecasts works, then this book is for you. While it is more technical, understanding the content of this book only requires (some) matrix algebra. Readers do not need to have a deep knowledge of statistics. The book is also freely available online in two different editions, which are based on two flavors of the free statistical computing platform R (R Core Team, 2022): the 2nd edition (https://otexts.com/fpp2/) uses base R and the forecast package (Hyndman et al., 2023), whereas the 3rd edition (https://otexts.com/fpp3/) uses the tidyverse (with a somewhat steeper learning curve than base R) and the newer fable package (O’Hara-Wild et al., 2020).
Shmueli and Lichtendahl (2018) is an introductory textbook that also focuses on forecasting with R, with many worked examples and exercises. There is also an analogous textbook (Shmueli, 2016), which instead uses Microsoft Excel and the XLMiner add-on.
Ord et al. (2017) is a software-agnostic textbook that is a little more verbose and business-user-friendly than Hyndman and Athanasopoulos (2021).
22.3 Technical material
If you want to get up to speed with the latest developments in forecasting, your best bet is to look at the International Journal of Forecasting (IJF; https://www.sciencedirect.com/journal/international-journal-of-forecasting), published by the International Institute of Forecasters (see below). Most articles are abstract but at the cutting edge of research in forecasting. The journal publishes regular special issues dedicated, for example, to the M forecasting competitions. (The similarly named Journal of Forecasting, https://onlinelibrary.wiley.com/journal/1099131x, focuses much more on forecasting other processes than demand, e.g., financial or macroeconomic time series.)
One recent paper published in the IJF is the comprehensive review article by Petropoulos, Apiletti, et al. (2022). Almost 80 worldwide experts pooled their knowledge to describe the current state of the art in forecasting, for demand as well as for other applications. The review includes ample references to recent academic papers. The article is published online at https://forecasting-encyclopedia.com/, and the authors plan on keeping this online version current.
Experts in Machine Learning applying their skills to forecasting are often unaware of the specific challenges in this sub-field. Hewamalage et al. (2022) provide a valuable introduction to the pitfalls in forecast evaluation, geared to this group.
22.4 Non-profit organizations
Various non-profit organizations have a mission to expand knowledge about forecasting.
The International Institute of Forecasters (IIF; https://forecasters.org/) is a global association of academics, practitioners, and consultants in forecasting. It brings together all sides involved in forecasting, allowing mutual learning. The IIF publishes two journals, the more practitioner-oriented Foresight: The International Journal of Applied Forecasting and the more academic International Journal of Forecasting (see above for both), and organizes an annual conference, the International Symposium on Forecasting (ISF; see below). The IIF also certifies courses in forecasting at various universities around the world.
The Centre for Marketing Analytics and Forecasting (CMAF; https://www.lancaster.ac.uk/lums/research/areas-of-expertise/centre-for-marketing-analytics-and-forecasting/) at Lancaster University Management School in the UK is the most important center of excellence in forecasting worldwide. It is an academic organization, but it offers consulting services and educates professionals in forecasting. It also provides very well-trained students for forecasting projects.
22.5 Events
The most relevant event in forecasting is the annual International Symposium on Forecasting (ISF; https://isf.forecasters.org/). This global conference is organized by the International Institute of Forecasters (IIF; see above). Attendees are an eclectic mix of academics and practitioners. There are dedicated practitioner tracks, but the academic tracks are also worth attending. Various workshops are offered before the conference as well.
22.6 Datasets
Kaggle (https://www.kaggle.com/) is a general data science platform that regularly organizes competitions, some of which are forecasting-themed. You can browse through the forums to find out how people address an ongoing or finished challenge, download data to try your algorithms, or even contact Kaggle to have them organize a forecasting competition using your data!
The Monash Time Series Forecasting Repository at https://forecastingdata.org/ (Godahewa et al., 2021) contains datasets of related time series, which are especially suitable for testing hierarchical forecasting techniques as in Chapter 13.
Many packages for the free statistical computing platform R (R Core Team, 2022) contain time series datasets. For instance, the fpp2 (Hyndman, 2020) and the fpp3 packages (Hyndman, 2023) contain the datasets used in the 2nd and 3rd edition of Hyndman and Athanasopoulos (2021) (see above), and more data can be found in the forecast (Hyndman et al., 2023), the fable (O’Hara-Wild et al., 2020) and the tsdl (Hyndman and Yang, 2023) packages. The Mcomp package (Hyndman, 2018) contains the datasets used in the M1 and the M3 competition. The data for the M5 competition can be found at https://www.kaggle.com/c/m5-forecasting-accuracy. (The recently concluded M6 competition used financial time series and is thus a little less helpful for demand forecasting.)
Many more datasets can be obtained from national statistical offices, or by a simple internet search. For instance, you can download time series of posts to the StackExchange network through the StackExchange Data Explorer (https://data.stackexchange.com/). Finally, if you need a very specific dataset, you can always ask at the OpenData StackExchange site (https://opendata.stackexchange.com/).
22.7 Online resources
Of course, most resources described above have an online presence, and we noted the URL wherever applicable. There are also a few purely online sites of interest to forecasters.
The CMAF mentioned above organizes a regular schedule of forecasting-themed live-streamed webinars, the Friday Forecasting Talks (FFTs; https://cmaf-fft.lp151.com/). Past webinars are archived on the CMAF YouTube channel (https://www.youtube.com/@lancastercmaf). The CMAF is publishing a video playlist on “Business Forecasting Principles”.
The IIF (see above) publishes a series of podcasts called “Forecasting Impact”, which features interviews with forecasting experts. Past episodes are archived at https://forecasters.org/publications/forecasting-impact-podcast/.
Finally, an invaluable resource for the practicing forecaster is CrossValidated (https://stats.stackexchange.com), a Q&A site for statistics with many forecasting questions (and answers). With over 200,000 questions, your specific forecasting question may already have a response, or you could start a new thread (just please read the site help first: https://stats.stackexchange.com/help).
Key takeaways
There are lots of resources to improve your forecasting – but don’t forget the other facets that make good forecasters (see Section 19.1).
Happy forecasting!