Course Overview
This course is a compact tour of how demand forecasting works in a supply chain setting. It is taught by Akshay Yewale, a supply chain practitioner with experience in FMCG, automotive, semiconductor and chemical companies. It focuses on process and judgment more than on software. You learn how to frame demand, how far ahead and how often to forecast, and how to decide whether a forecast is any good.
Forecasting is the starting point of supply chain planning, so the course keeps returning to the link between forecast quality and business decisions. Its ideas include collaborating on the forecast instead of treating it as a solo number, separating baseline demand from promotional lift, and cleaning outliers. It also covers the organizational side: how a demand review meeting is run, what goes on its agenda, and which metrics to bring.
What You Will Learn
- How to define demand so that it reflects real customer need, not just recorded sales
- How to set forecast horizon, period and review frequency
- Separating base sales from incremental sales
- Correcting outliers with winsorization, with a practice assignment
- Cannibalization and halo effects, and the different types of cannibalization
- Segmenting demand, SKUs and models so each item gets a suitable method
- Accuracy measures including RMSE, MAE, MAPE and bias
- Training and testing datasets, model initialization and data leakage
- How much history is too much when training a model
- Choosing a model, building an exception review report, and running forecast reviews
- Traditional statistical models compared with machine learning models
Course Structure
The 33 lectures are spread over 8 sections. Most run between 3 and 12 minutes, with one article on choosing the right forecasting model.
- Orientation: platform navigation, how to use the course, instructor introduction and content overview
- Principles and demand definition: forecasting principles, redefining demand, collaboration, horizon and review cadence
- Data preparation: base and incremental sales, winsorization, cannibalization and halo effect
- Segmentation and testing: demand segmentation, whether model testing is sufficient, and accuracy metrics
- Modeling practice: training and testing data, data leakage, why forecasts miss, and model selection methodology
- Review process: exception reports, forecast review, and the demand review agenda, metrics, inputs, outputs and tips
- Model families: traditional statistical models, machine learning models and a comparison of the two
Who Is This Course For?
The course is aimed at demand planners, data scientists and students of forecasting. It suits you if you are new to demand planning and want a map of the whole process. It also suits you if you work near forecasting, for example in supply chain, sales operations or inventory, and need to speak the same language as the planners.
Look elsewhere if you want to build forecasting models step by step in code or a spreadsheet. The course is mainly conceptual and procedural. People who already run demand reviews daily will probably recognize much of the material.
Format & Time Commitment
The course is self-paced video, about 3.5 hours in total, with assignments. It has closed captions, can be watched on mobile and TV, and comes with lifetime access. Most learners can finish it in a few evenings or a long weekend. The instructor says he updates the course when new practices come up and takes topic requests through reviews.
Pros and Cons
Pros
- Covers the full planning loop, from defining demand to running the review meeting, not just the statistics
- Short lectures make it easy to study in small sessions
- Covers real-world issues such as data leakage, cannibalization and outlier handling
- Needs only basic Excel
- Includes assignments and a completion certificate
Cons
- The 3.8/5 rating from 79 reviews is middling. One reviewer complained about audio quality and content, and another found the delivery uneven.
- At around three hours for 33 topics, many subjects get only a brief introduction. Expect awareness, not mastery.
- The listing shows a last update of 8/2023, so the machine learning material may feel less current.
- A Udemy completion certificate shows you finished the course. It is not a recognized professional credential.
- It leans toward concepts over hands-on model building, so you will need another resource to practice implementation.
FAQ
Do I need a forecasting or statistics background?
No. The listing asks only for basic Excel.
Will I get a certificate?
Yes, a certificate of completion. It is useful as a record of learning but does not carry formal accreditation.
Is this course technical enough for data scientists?
It covers metrics, train/test splits and model comparison, but at an overview level. Data scientists will likely get the most value from the business-process lectures.
What else could I consider?
The same instructor offers a longer course on end-to-end demand forecasting. Other Udemy options on the same topic include one focused on forecasting KPIs and another that builds forecasts in Excel.
If the topic list matches your needs, check the course page on Udemy for the sample lectures and current details before you enroll.
